godotengine
/
godot
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
godot/gles_builders.py

745 lines
26 KiB
Python
Raw Normal View History

Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
"""Functions used to generate source files during build time
All such functions are invoked in a subprocess on Windows to prevent build flakiness.
"""
from platform_methods import subprocess_main
class LegacyGLHeaderStruct:
def __init__(self):
self.vertex_lines = []
self.fragment_lines = []
self.uniforms = []
self.attributes = []
self.feedbacks = []
self.fbos = []
self.conditionals = []
self.enums = {}
self.texunits = []
self.texunit_names = []
self.ubos = []
self.ubo_names = []
self.vertex_included_files = []
self.fragment_included_files = []
self.reading = ""
self.line_offset = 0
self.vertex_offset = 0
self.fragment_offset = 0
def include_file_in_legacygl_header(filename, header_data, depth):
fs = open(filename, "r")
line = fs.readline()
while line:
if line.find("[vertex]") != -1:
header_data.reading = "vertex"
line = fs.readline()
header_data.line_offset += 1
header_data.vertex_offset = header_data.line_offset
continue
if line.find("[fragment]") != -1:
header_data.reading = "fragment"
line = fs.readline()
header_data.line_offset += 1
header_data.fragment_offset = header_data.line_offset
continue
while line.find("#include ") != -1:
includeline = line.replace("#include ", "").strip()[1:-1]
import os.path
included_file = os.path.relpath(os.path.dirname(filename) + "/" + includeline)
if not included_file in header_data.vertex_included_files and header_data.reading == "vertex":
header_data.vertex_included_files += [included_file]
Dont use equality operators with None singleton in python files
2018-10-26 23:18:15 +00:00
if include_file_in_legacygl_header(included_file, header_data, depth + 1) is None:
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
print("Error in file '" + filename + "': #include " + includeline + "could not be found!")
elif not included_file in header_data.fragment_included_files and header_data.reading == "fragment":
header_data.fragment_included_files += [included_file]
Dont use equality operators with None singleton in python files
2018-10-26 23:18:15 +00:00
if include_file_in_legacygl_header(included_file, header_data, depth + 1) is None:
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
print("Error in file '" + filename + "': #include " + includeline + "could not be found!")
line = fs.readline()
-Lightmap and lightmap capture support for GLES2 -Added hint to not show some properties when running on low end gfx
2018-09-28 23:32:40 +00:00
if line.find("#ifdef ") != -1:
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if line.find("#ifdef ") != -1:
ifdefline = line.replace("#ifdef ", "").strip()
if line.find("_EN_") != -1:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
enumbase = ifdefline[: ifdefline.find("_EN_")]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
ifdefline = ifdefline.replace("_EN_", "_")
line = line.replace("_EN_", "_")
if enumbase not in header_data.enums:
header_data.enums[enumbase] = []
if ifdefline not in header_data.enums[enumbase]:
header_data.enums[enumbase].append(ifdefline)
elif not ifdefline in header_data.conditionals:
header_data.conditionals += [ifdefline]
if line.find("uniform") != -1 and line.lower().find("texunit:") != -1:
# texture unit
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
texunitstr = line[line.find(":") + 1 :].strip()
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if texunitstr == "auto":
texunit = "-1"
else:
texunit = str(int(texunitstr))
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
uline = line[: line.lower().find("//")]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
uline = uline.replace("uniform", "")
uline = uline.replace("highp", "")
uline = uline.replace(";", "")
lines = uline.split(",")
for x in lines:
x = x.strip()
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
x = x[x.rfind(" ") + 1 :]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if x.find("[") != -1:
# unfiorm array
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
x = x[: x.find("[")]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if not x in header_data.texunit_names:
header_data.texunits += [(x, texunit)]
header_data.texunit_names += [x]
elif line.find("uniform") != -1 and line.lower().find("ubo:") != -1:
# uniform buffer object
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
ubostr = line[line.find(":") + 1 :].strip()
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
ubo = str(int(ubostr))
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
uline = line[: line.lower().find("//")]
uline = uline[uline.find("uniform") + len("uniform") :]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
uline = uline.replace("highp", "")
uline = uline.replace(";", "")
uline = uline.replace("{", "").strip()
lines = uline.split(",")
for x in lines:
x = x.strip()
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
x = x[x.rfind(" ") + 1 :]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if x.find("[") != -1:
# unfiorm array
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
x = x[: x.find("[")]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if not x in header_data.ubo_names:
header_data.ubos += [(x, ubo)]
header_data.ubo_names += [x]
elif line.find("uniform") != -1 and line.find("{") == -1 and line.find(";") != -1:
uline = line.replace("uniform", "")
uline = uline.replace(";", "")
lines = uline.split(",")
for x in lines:
x = x.strip()
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
x = x[x.rfind(" ") + 1 :]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if x.find("[") != -1:
# unfiorm array
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
x = x[: x.find("[")]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if not x in header_data.uniforms:
header_data.uniforms += [x]
if line.strip().find("attribute ") == 0 and line.find("attrib:") != -1:
uline = line.replace("in ", "")
uline = uline.replace("attribute ", "")
uline = uline.replace("highp ", "")
uline = uline.replace(";", "")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
uline = uline[uline.find(" ") :].strip()
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if uline.find("//") != -1:
name, bind = uline.split("//")
if bind.find("attrib:") != -1:
name = name.strip()
bind = bind.replace("attrib:", "").strip()
header_data.attributes += [(name, bind)]
if line.strip().find("out ") == 0 and line.find("tfb:") != -1:
uline = line.replace("out ", "")
uline = uline.replace("highp ", "")
uline = uline.replace(";", "")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
uline = uline[uline.find(" ") :].strip()
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
if uline.find("//") != -1:
name, bind = uline.split("//")
if bind.find("tfb:") != -1:
name = name.strip()
bind = bind.replace("tfb:", "").strip()
header_data.feedbacks += [(name, bind)]
line = line.replace("\r", "")
line = line.replace("\n", "")
if header_data.reading == "vertex":
header_data.vertex_lines += [line]
if header_data.reading == "fragment":
header_data.fragment_lines += [line]
line = fs.readline()
header_data.line_offset += 1
fs.close()
return header_data
def build_legacygl_header(filename, include, class_suffix, output_attribs, gles2=False):
header_data = LegacyGLHeaderStruct()
include_file_in_legacygl_header(filename, header_data, 0)
out_file = filename + ".gen.h"
fd = open(out_file, "w")
enum_constants = []
fd.write("/* WARNING, THIS FILE WAS GENERATED, DO NOT EDIT */\n")
out_file_base = out_file
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
out_file_base = out_file_base[out_file_base.rfind("/") + 1 :]
out_file_base = out_file_base[out_file_base.rfind("\\") + 1 :]
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
out_file_ifdef = out_file_base.replace(".", "_").upper()
fd.write("#ifndef " + out_file_ifdef + class_suffix + "_120\n")
fd.write("#define " + out_file_ifdef + class_suffix + "_120\n")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
out_file_class = (
out_file_base.replace(".glsl.gen.h", "").title().replace("_", "").replace(".", "") + "Shader" + class_suffix
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\n\n")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('#include "' + include + '"\n\n\n')
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("class " + out_file_class + " : public Shader" + class_suffix + " {\n\n")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t virtual String get_shader_name() const { return "' + out_file_class + '"; }\n')
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("public:\n\n")
if header_data.conditionals:
fd.write("\tenum Conditionals {\n")
for x in header_data.conditionals:
fd.write("\t\t" + x.upper() + ",\n")
fd.write("\t};\n\n")
if header_data.uniforms:
fd.write("\tenum Uniforms {\n")
for x in header_data.uniforms:
fd.write("\t\t" + x.upper() + ",\n")
fd.write("\t};\n\n")
fd.write("\t_FORCE_INLINE_ int get_uniform(Uniforms p_uniform) const { return _get_uniform(p_uniform); }\n\n")
if header_data.conditionals:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"\t_FORCE_INLINE_ void set_conditional(Conditionals p_conditional,bool p_enable) { _set_conditional(p_conditional,p_enable); }\n\n"
)
-Rewrote GLES2 lighting and shadows and optimized state changes, did many optimizations, added vertex lighting. -Did some fixes to GLES3 too
2018-09-23 15:12:30 +00:00
fd.write("\t#ifdef DEBUG_ENABLED\n ")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"\t#define _FU if (get_uniform(p_uniform)<0) return; if (!is_version_valid()) return; ERR_FAIL_COND( get_active()!=this ); \n\n "
)
Remove trailing whitespace With `sed -i $(rg -l '[[:blank:]]*$' -g'!thirdparty') -e 's/[[:blank:]]*$//g'` (+ manual revert of some thirdparty code under `platform/android`).
2018-11-20 10:14:07 +00:00
fd.write("\t#else\n ")
-Rewrote GLES2 lighting and shadows and optimized state changes, did many optimizations, added vertex lighting. -Did some fixes to GLES3 too
2018-09-23 15:12:30 +00:00
fd.write("\t#define _FU if (get_uniform(p_uniform)<0) return; \n\n ")
Remove trailing whitespace With `sed -i $(rg -l '[[:blank:]]*$' -g'!thirdparty') -e 's/[[:blank:]]*$//g'` (+ manual revert of some thirdparty code under `platform/android`).
2018-11-20 10:14:07 +00:00
fd.write("\t#endif\n")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, float p_value) { _FU glUniform1f(get_uniform(p_uniform),p_value); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, double p_value) { _FU glUniform1f(get_uniform(p_uniform),p_value); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, uint8_t p_value) { _FU glUniform1i(get_uniform(p_uniform),p_value); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, int8_t p_value) { _FU glUniform1i(get_uniform(p_uniform),p_value); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, uint16_t p_value) { _FU glUniform1i(get_uniform(p_uniform),p_value); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, int16_t p_value) { _FU glUniform1i(get_uniform(p_uniform),p_value); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, uint32_t p_value) { _FU glUniform1i(get_uniform(p_uniform),p_value); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, int32_t p_value) { _FU glUniform1i(get_uniform(p_uniform),p_value); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, const Color& p_color) { _FU GLfloat col[4]={p_color.r,p_color.g,p_color.b,p_color.a}; glUniform4fv(get_uniform(p_uniform),1,col); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, const Vector2& p_vec2) { _FU GLfloat vec2[2]={p_vec2.x,p_vec2.y}; glUniform2fv(get_uniform(p_uniform),1,vec2); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, const Size2i& p_vec2) { _FU GLint vec2[2]={p_vec2.x,p_vec2.y}; glUniform2iv(get_uniform(p_uniform),1,vec2); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, const Vector3& p_vec3) { _FU GLfloat vec3[3]={p_vec3.x,p_vec3.y,p_vec3.z}; glUniform3fv(get_uniform(p_uniform),1,vec3); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, float p_a, float p_b) { _FU glUniform2f(get_uniform(p_uniform),p_a,p_b); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, float p_a, float p_b, float p_c) { _FU glUniform3f(get_uniform(p_uniform),p_a,p_b,p_c); }\n\n"
)
fd.write(
"\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, float p_a, float p_b, float p_c, float p_d) { _FU glUniform4f(get_uniform(p_uniform),p_a,p_b,p_c,p_d); }\n\n"
)
fd.write(
"""\t_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, const Transform& p_transform) { _FU
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
const Transform &tr = p_transform;
GLfloat matrix[16]={ /* build a 16x16 matrix */
tr.basis.elements[0][0],
tr.basis.elements[1][0],
tr.basis.elements[2][0],
0,
tr.basis.elements[0][1],
tr.basis.elements[1][1],
tr.basis.elements[2][1],
0,
tr.basis.elements[0][2],
tr.basis.elements[1][2],
tr.basis.elements[2][2],
0,
tr.origin.x,
tr.origin.y,
tr.origin.z,
1
};
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
glUniformMatrix4fv(get_uniform(p_uniform),1,false,matrix);
}
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
"""
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"""_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, const Transform2D& p_transform) { _FU
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
const Transform2D &tr = p_transform;
GLfloat matrix[16]={ /* build a 16x16 matrix */
tr.elements[0][0],
tr.elements[0][1],
0,
0,
tr.elements[1][0],
tr.elements[1][1],
0,
0,
0,
0,
1,
0,
tr.elements[2][0],
tr.elements[2][1],
0,
1
};
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
glUniformMatrix4fv(get_uniform(p_uniform),1,false,matrix);
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
}
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
"""
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"""_FORCE_INLINE_ void set_uniform(Uniforms p_uniform, const CameraMatrix& p_matrix) { _FU
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
GLfloat matrix[16];
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
for (int i=0;i<4;i++) {
for (int j=0;j<4;j++) {
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
matrix[i*4+j]=p_matrix.matrix[i][j];
}
}
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
glUniformMatrix4fv(get_uniform(p_uniform),1,false,matrix);
}
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
"""
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\n\n#undef _FU\n\n\n")
fd.write("\tvirtual void init() {\n\n")
enum_value_count = 0
if header_data.enums:
fd.write("\t\t//Written using math, given nonstandarity of 64 bits integer constants..\n")
fd.write("\t\tstatic const Enum _enums[]={\n")
bitofs = len(header_data.conditionals)
enum_vals = []
for xv in header_data.enums:
x = header_data.enums[xv]
bits = 1
amt = len(x)
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
while 2 ** bits < amt:
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
bits += 1
strs = "{"
for i in range(amt):
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
strs += '"#define ' + x[i] + '\\n",'
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
c = {}
c["set_mask"] = "uint64_t(" + str(i) + ")<<" + str(bitofs)
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
c["clear_mask"] = (
"((uint64_t(1)<<40)-1) ^ (((uint64_t(1)<<" + str(bits) + ") - 1)<<" + str(bitofs) + ")"
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
enum_vals.append(c)
enum_constants.append(x[i])
strs += "NULL}"
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"\t\t\t{(uint64_t(1<<" + str(bits) + ")-1)<<" + str(bitofs) + "," + str(bitofs) + "," + strs + "},\n"
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
bitofs += bits
fd.write("\t\t};\n\n")
fd.write("\t\tstatic const EnumValue _enum_values[]={\n")
enum_value_count = len(enum_vals)
for x in enum_vals:
fd.write("\t\t\t{" + x["set_mask"] + "," + x["clear_mask"] + "},\n")
fd.write("\t\t};\n\n")
conditionals_found = []
if header_data.conditionals:
fd.write("\t\tstatic const char* _conditional_strings[]={\n")
if header_data.conditionals:
for x in header_data.conditionals:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\t\t"#define ' + x + '\\n",\n')
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
conditionals_found.append(x)
fd.write("\t\t};\n\n")
else:
fd.write("\t\tstatic const char **_conditional_strings=NULL;\n")
if header_data.uniforms:
fd.write("\t\tstatic const char* _uniform_strings[]={\n")
if header_data.uniforms:
for x in header_data.uniforms:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\t\t"' + x + '",\n')
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\t\t};\n\n")
else:
fd.write("\t\tstatic const char **_uniform_strings=NULL;\n")
if output_attribs:
if header_data.attributes:
fd.write("\t\tstatic AttributePair _attribute_pairs[]={\n")
for x in header_data.attributes:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\t\t{"' + x[0] + '",' + x[1] + "},\n")
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\t\t};\n\n")
else:
fd.write("\t\tstatic AttributePair *_attribute_pairs=NULL;\n")
feedback_count = 0
if not gles2 and len(header_data.feedbacks):
fd.write("\t\tstatic const Feedback _feedbacks[]={\n")
for x in header_data.feedbacks:
name = x[0]
cond = x[1]
if cond in conditionals_found:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\t\t{"' + name + '",' + str(conditionals_found.index(cond)) + "},\n")
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
else:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\t\t{"' + name + '",-1},\n')
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
feedback_count += 1
fd.write("\t\t};\n\n")
else:
if gles2:
pass
else:
fd.write("\t\tstatic const Feedback* _feedbacks=NULL;\n")
if header_data.texunits:
fd.write("\t\tstatic TexUnitPair _texunit_pairs[]={\n")
for x in header_data.texunits:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\t\t{"' + x[0] + '",' + x[1] + "},\n")
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\t\t};\n\n")
else:
fd.write("\t\tstatic TexUnitPair *_texunit_pairs=NULL;\n")
if not gles2 and header_data.ubos:
fd.write("\t\tstatic UBOPair _ubo_pairs[]={\n")
for x in header_data.ubos:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\t\t{"' + x[0] + '",' + x[1] + "},\n")
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\t\t};\n\n")
else:
if gles2:
pass
else:
fd.write("\t\tstatic UBOPair *_ubo_pairs=NULL;\n")
fd.write("\t\tstatic const char _vertex_code[]={\n")
for x in header_data.vertex_lines:
for c in x:
fd.write(str(ord(c)) + ",")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(str(ord("\n")) + ",")
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\t\t0};\n\n")
fd.write("\t\tstatic const int _vertex_code_start=" + str(header_data.vertex_offset) + ";\n")
fd.write("\t\tstatic const char _fragment_code[]={\n")
for x in header_data.fragment_lines:
for c in x:
fd.write(str(ord(c)) + ",")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(str(ord("\n")) + ",")
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\t\t0};\n\n")
fd.write("\t\tstatic const int _fragment_code_start=" + str(header_data.fragment_offset) + ";\n")
if output_attribs:
if gles2:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"\t\tsetup(_conditional_strings,"
+ str(len(header_data.conditionals))
+ ",_uniform_strings,"
+ str(len(header_data.uniforms))
+ ",_attribute_pairs,"
+ str(len(header_data.attributes))
+ ", _texunit_pairs,"
+ str(len(header_data.texunits))
+ ",_vertex_code,_fragment_code,_vertex_code_start,_fragment_code_start);\n"
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
else:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"\t\tsetup(_conditional_strings,"
+ str(len(header_data.conditionals))
+ ",_uniform_strings,"
+ str(len(header_data.uniforms))
+ ",_attribute_pairs,"
+ str(len(header_data.attributes))
+ ", _texunit_pairs,"
+ str(len(header_data.texunits))
+ ",_ubo_pairs,"
+ str(len(header_data.ubos))
+ ",_feedbacks,"
+ str(feedback_count)
+ ",_vertex_code,_fragment_code,_vertex_code_start,_fragment_code_start);\n"
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
else:
if gles2:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"\t\tsetup(_conditional_strings,"
+ str(len(header_data.conditionals))
+ ",_uniform_strings,"
+ str(len(header_data.uniforms))
+ ",_texunit_pairs,"
+ str(len(header_data.texunits))
+ ",_enums,"
+ str(len(header_data.enums))
+ ",_enum_values,"
+ str(enum_value_count)
+ ",_vertex_code,_fragment_code,_vertex_code_start,_fragment_code_start);\n"
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
else:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(
"\t\tsetup(_conditional_strings,"
+ str(len(header_data.conditionals))
+ ",_uniform_strings,"
+ str(len(header_data.uniforms))
+ ",_texunit_pairs,"
+ str(len(header_data.texunits))
+ ",_enums,"
+ str(len(header_data.enums))
+ ",_enum_values,"
+ str(enum_value_count)
+ ",_ubo_pairs,"
+ str(len(header_data.ubos))
+ ",_feedbacks,"
+ str(feedback_count)
+ ",_vertex_code,_fragment_code,_vertex_code_start,_fragment_code_start);\n"
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
fd.write("\t}\n\n")
if enum_constants:
fd.write("\tenum EnumConditionals {\n")
for x in enum_constants:
fd.write("\t\t" + x.upper() + ",\n")
fd.write("\t};\n\n")
fd.write("\tvoid set_enum_conditional(EnumConditionals p_cond) { _set_enum_conditional(p_cond); }\n")
fd.write("};\n\n")
fd.write("#endif\n\n")
fd.close()
def build_gles2_headers(target, source, env):
for x in source:
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
build_legacygl_header(
str(x), include="drivers/gles2/shader_gles2.h", class_suffix="GLES2", output_attribs=True, gles2=True
)
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
class RDHeaderStruct:
def __init__(self):
self.vertex_lines = []
self.fragment_lines = []
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
self.compute_lines = []
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
self.vertex_included_files = []
self.fragment_included_files = []
self.reading = ""
self.line_offset = 0
self.vertex_offset = 0
self.fragment_offset = 0
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
self.compute_offset = 0
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
def include_file_in_rd_header(filename, header_data, depth):
fs = open(filename, "r")
line = fs.readline()
while line:
if line.find("[vertex]") != -1:
header_data.reading = "vertex"
line = fs.readline()
header_data.line_offset += 1
header_data.vertex_offset = header_data.line_offset
continue
if line.find("[fragment]") != -1:
header_data.reading = "fragment"
line = fs.readline()
header_data.line_offset += 1
header_data.fragment_offset = header_data.line_offset
continue
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
if line.find("[compute]") != -1:
header_data.reading = "compute"
line = fs.readline()
header_data.line_offset += 1
header_data.compute_offset = header_data.line_offset
continue
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
while line.find("#include ") != -1:
includeline = line.replace("#include ", "").strip()[1:-1]
import os.path
included_file = os.path.relpath(os.path.dirname(filename) + "/" + includeline)
if not included_file in header_data.vertex_included_files and header_data.reading == "vertex":
header_data.vertex_included_files += [included_file]
if include_file_in_rd_header(included_file, header_data, depth + 1) is None:
print("Error in file '" + filename + "': #include " + includeline + "could not be found!")
elif not included_file in header_data.fragment_included_files and header_data.reading == "fragment":
header_data.fragment_included_files += [included_file]
if include_file_in_rd_header(included_file, header_data, depth + 1) is None:
print("Error in file '" + filename + "': #include " + includeline + "could not be found!")
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
elif not included_file in header_data.compute_included_files and header_data.reading == "compute":
header_data.compute_included_files += [included_file]
if include_file_in_rd_header(included_file, header_data, depth + 1) is None:
print("Error in file '" + filename + "': #include " + includeline + "could not be found!")
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
line = fs.readline()
line = line.replace("\r", "")
line = line.replace("\n", "")
if header_data.reading == "vertex":
header_data.vertex_lines += [line]
if header_data.reading == "fragment":
header_data.fragment_lines += [line]
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
if header_data.reading == "compute":
header_data.compute_lines += [line]
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
line = fs.readline()
header_data.line_offset += 1
fs.close()
return header_data
Better format generated shader headers
2019-09-27 11:34:26 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
def build_rd_header(filename):
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
header_data = RDHeaderStruct()
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
include_file_in_rd_header(filename, header_data, 0)
out_file = filename + ".gen.h"
fd = open(out_file, "w")
enum_constants = []
fd.write("/* WARNING, THIS FILE WAS GENERATED, DO NOT EDIT */\n")
out_file_base = out_file
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
out_file_base = out_file_base[out_file_base.rfind("/") + 1 :]
out_file_base = out_file_base[out_file_base.rfind("\\") + 1 :]
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
out_file_ifdef = out_file_base.replace(".", "_").upper()
fd.write("#ifndef " + out_file_ifdef + "_RD\n")
fd.write("#define " + out_file_ifdef + "_RD\n")
out_file_class = out_file_base.replace(".glsl.gen.h", "").title().replace("_", "").replace(".", "") + "ShaderRD"
Better format generated shader headers
2019-09-27 11:34:26 +00:00
fd.write("\n")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('#include "servers/rendering/rasterizer_rd/shader_rd.h"\n\n')
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
fd.write("class " + out_file_class + " : public ShaderRD {\n\n")
fd.write("public:\n\n")
Better format generated shader headers
2019-09-27 11:34:26 +00:00
fd.write("\t" + out_file_class + "() {\n\n")
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
if len(header_data.compute_lines):
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
Better format generated shader headers
2019-09-27 11:34:26 +00:00
fd.write("\t\tstatic const char _compute_code[] = {\n")
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
for x in header_data.compute_lines:
for c in x:
fd.write(str(ord(c)) + ",")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(str(ord("\n")) + ",")
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
fd.write("\t\t0};\n\n")
Better format generated shader headers
2019-09-27 11:34:26 +00:00
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\tsetup(nullptr, nullptr, _compute_code, "' + out_file_class + '");\n')
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
fd.write("\t}\n")
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
else:
Better format generated shader headers
2019-09-27 11:34:26 +00:00
fd.write("\t\tstatic const char _vertex_code[] = {\n")
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
for x in header_data.vertex_lines:
for c in x:
fd.write(str(ord(c)) + ",")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(str(ord("\n")) + ",")
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
fd.write("\t\t0};\n\n")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
fd.write("\t\tstatic const char _fragment_code[]={\n")
for x in header_data.fragment_lines:
for c in x:
Better format generated shader headers
2019-09-27 11:34:26 +00:00
fd.write(str(ord(c)) + ",")
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write(str(ord("\n")) + ",")
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
fd.write("\t\t0};\n\n")
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
fd.write('\t\tsetup(_vertex_code, _fragment_code, nullptr, "' + out_file_class + '");\n')
Better format generated shader headers
2019-09-27 11:34:26 +00:00
fd.write("\t}\n")
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
fd.write("};\n\n")
Better format generated shader headers
2019-09-27 11:34:26 +00:00
fd.write("#endif\n")
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
fd.close()
def build_rd_headers(target, source, env):
for x in source:
build_rd_header(str(x))
SCons: Format buildsystem files with psf/black Configured for a max line length of 120 characters. psf/black is very opinionated and purposely doesn't leave much room for configuration. The output is mostly OK so that should be fine for us, but some things worth noting: - Manually wrapped strings will be reflowed, so by using a line length of 120 for the sake of preserving readability for our long command calls, it also means that some manually wrapped strings are back on the same line and should be manually merged again. - Code generators using string concatenation extensively look awful, since black puts each operand on a single line. We need to refactor these generators to use more pythonic string formatting, for which many options are available (`%`, `format` or f-strings). - CI checks and a pre-commit hook will be added to ensure that future buildsystem changes are well-formatted.
2020-03-30 06:28:32 +00:00
if __name__ == "__main__":
Running builder (content generator) functions in subprocesses on Windows - Refactored all builder (make_*) functions into separate Python modules along to the build tree - Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere - Introduced stub to use the builders module as a stand alone script and invoke a selected function There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp) on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky builds. Running all such content generators in a new subprocess instead of directly inside the build script works around the issue. Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle. Suggested workaround did not fully work either. Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of running a cross-compilation on Windows they would still be used, but likely it will not happen in practice. What counts is that the build itself is running on which platform, not the target platform. Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
2018-03-17 22:23:55 +00:00
subprocess_main(globals())