godot/modules/mono/build_scripts/build_assemblies.py

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

436 lines
15 KiB
Python
Raw Permalink Normal View History

C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
#!/usr/bin/python3
import os
import os.path
import shlex
import subprocess
from dataclasses import dataclass
from typing import List, Optional
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
def find_dotnet_cli():
if os.name == "nt":
for hint_dir in os.environ["PATH"].split(os.pathsep):
hint_dir = hint_dir.strip('"')
hint_path = os.path.join(hint_dir, "dotnet")
if os.path.isfile(hint_path) and os.access(hint_path, os.X_OK):
return hint_path
if os.path.isfile(hint_path + ".exe") and os.access(hint_path + ".exe", os.X_OK):
return hint_path + ".exe"
else:
for hint_dir in os.environ["PATH"].split(os.pathsep):
hint_dir = hint_dir.strip('"')
hint_path = os.path.join(hint_dir, "dotnet")
if os.path.isfile(hint_path) and os.access(hint_path, os.X_OK):
return hint_path
def find_msbuild_standalone_windows():
msbuild_tools_path = find_msbuild_tools_path_reg()
if msbuild_tools_path:
return os.path.join(msbuild_tools_path, "MSBuild.exe")
return None
def find_msbuild_mono_windows(mono_prefix):
assert mono_prefix is not None
mono_bin_dir = os.path.join(mono_prefix, "bin")
msbuild_mono = os.path.join(mono_bin_dir, "msbuild.bat")
if os.path.isfile(msbuild_mono):
return msbuild_mono
return None
def find_msbuild_mono_unix():
import sys
hint_dirs = []
if sys.platform == "darwin":
hint_dirs[:0] = [
"/Library/Frameworks/Mono.framework/Versions/Current/bin",
"/usr/local/var/homebrew/linked/mono/bin",
]
for hint_dir in hint_dirs:
hint_path = os.path.join(hint_dir, "msbuild")
if os.path.isfile(hint_path):
return hint_path
elif os.path.isfile(hint_path + ".exe"):
return hint_path + ".exe"
for hint_dir in os.environ["PATH"].split(os.pathsep):
hint_dir = hint_dir.strip('"')
hint_path = os.path.join(hint_dir, "msbuild")
if os.path.isfile(hint_path) and os.access(hint_path, os.X_OK):
return hint_path
if os.path.isfile(hint_path + ".exe") and os.access(hint_path + ".exe", os.X_OK):
return hint_path + ".exe"
return None
def find_msbuild_tools_path_reg():
import subprocess
program_files = os.getenv("PROGRAMFILES(X86)")
if not program_files:
program_files = os.getenv("PROGRAMFILES")
vswhere = os.path.join(program_files, "Microsoft Visual Studio", "Installer", "vswhere.exe")
vswhere_args = ["-latest", "-products", "*", "-requires", "Microsoft.Component.MSBuild"]
try:
lines = subprocess.check_output([vswhere] + vswhere_args).splitlines()
for line in lines:
parts = line.decode("utf-8").split(":", 1)
if len(parts) < 2 or parts[0] != "installationPath":
continue
val = parts[1].strip()
if not val:
raise ValueError("Value of `installationPath` entry is empty")
# Since VS2019, the directory is simply named "Current"
msbuild_dir = os.path.join(val, "MSBuild", "Current", "Bin")
if os.path.isdir(msbuild_dir):
return msbuild_dir
# Directory name "15.0" is used in VS 2017
return os.path.join(val, "MSBuild", "15.0", "Bin")
raise ValueError("Cannot find `installationPath` entry")
except ValueError as e:
print("Error reading output from vswhere: " + str(e))
except OSError:
pass # Fine, vswhere not found
except (subprocess.CalledProcessError, OSError):
pass
@dataclass
class ToolsLocation:
dotnet_cli: str = ""
msbuild_standalone: str = ""
msbuild_mono: str = ""
mono_bin_dir: str = ""
def find_any_msbuild_tool(mono_prefix):
# Preference order: dotnet CLI > Standalone MSBuild > Mono's MSBuild
# Find dotnet CLI
dotnet_cli = find_dotnet_cli()
if dotnet_cli:
return ToolsLocation(dotnet_cli=dotnet_cli)
# Find standalone MSBuild
if os.name == "nt":
msbuild_standalone = find_msbuild_standalone_windows()
if msbuild_standalone:
return ToolsLocation(msbuild_standalone=msbuild_standalone)
if mono_prefix:
# Find Mono's MSBuild
if os.name == "nt":
msbuild_mono = find_msbuild_mono_windows(mono_prefix)
if msbuild_mono:
return ToolsLocation(msbuild_mono=msbuild_mono)
else:
msbuild_mono = find_msbuild_mono_unix()
if msbuild_mono:
return ToolsLocation(msbuild_mono=msbuild_mono)
return None
def run_msbuild(tools: ToolsLocation, sln: str, chdir_to: str, msbuild_args: Optional[List[str]] = None):
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
using_msbuild_mono = False
# Preference order: dotnet CLI > Standalone MSBuild > Mono's MSBuild
if tools.dotnet_cli:
args = [tools.dotnet_cli, "msbuild"]
elif tools.msbuild_standalone:
args = [tools.msbuild_standalone]
elif tools.msbuild_mono:
args = [tools.msbuild_mono]
using_msbuild_mono = True
else:
raise RuntimeError("Path to MSBuild or dotnet CLI not provided.")
args += [sln]
if msbuild_args:
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
args += msbuild_args
print("Running MSBuild: ", " ".join(shlex.quote(arg) for arg in args), flush=True)
msbuild_env = os.environ.copy()
# Needed when running from Developer Command Prompt for VS
if "PLATFORM" in msbuild_env:
del msbuild_env["PLATFORM"]
if using_msbuild_mono:
# The (Csc/Vbc/Fsc)ToolExe environment variables are required when
# building with Mono's MSBuild. They must point to the batch files
# in Mono's bin directory to make sure they are executed with Mono.
msbuild_env.update(
{
"CscToolExe": os.path.join(tools.mono_bin_dir, "csc.bat"),
"VbcToolExe": os.path.join(tools.mono_bin_dir, "vbc.bat"),
"FscToolExe": os.path.join(tools.mono_bin_dir, "fsharpc.bat"),
}
)
# We want to control cwd when running msbuild, because that's where the search for global.json begins.
return subprocess.call(args, env=msbuild_env, cwd=chdir_to)
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
def build_godot_api(msbuild_tool, module_dir, output_dir, push_nupkgs_local, precision, no_deprecated):
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
target_filenames = [
"GodotSharp.dll",
"GodotSharp.pdb",
"GodotSharp.xml",
"GodotSharpEditor.dll",
"GodotSharpEditor.pdb",
"GodotSharpEditor.xml",
"GodotPlugins.dll",
"GodotPlugins.pdb",
"GodotPlugins.runtimeconfig.json",
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
]
for build_config in ["Debug", "Release"]:
editor_api_dir = os.path.join(output_dir, "GodotSharp", "Api", build_config)
targets = [os.path.join(editor_api_dir, filename) for filename in target_filenames]
args = ["/restore", "/t:Build", "/p:Configuration=" + build_config, "/p:NoWarn=1591"]
if push_nupkgs_local:
args += ["/p:ClearNuGetLocalCache=true", "/p:PushNuGetToLocalSource=" + push_nupkgs_local]
if precision == "double":
2022-09-02 17:08:40 +00:00
args += ["/p:GodotFloat64=true"]
if no_deprecated:
args += ["/p:GodotNoDeprecated=true"]
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
sln = os.path.join(module_dir, "glue/GodotSharp/GodotSharp.sln")
exit_code = run_msbuild(msbuild_tool, sln=sln, chdir_to=module_dir, msbuild_args=args)
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
if exit_code != 0:
return exit_code
# Copy targets
core_src_dir = os.path.abspath(os.path.join(sln, os.pardir, "GodotSharp", "bin", build_config))
editor_src_dir = os.path.abspath(os.path.join(sln, os.pardir, "GodotSharpEditor", "bin", build_config))
2022-02-27 20:57:50 +00:00
plugins_src_dir = os.path.abspath(os.path.join(sln, os.pardir, "GodotPlugins", "bin", build_config, "net6.0"))
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
if not os.path.isdir(editor_api_dir):
assert not os.path.isfile(editor_api_dir)
os.makedirs(editor_api_dir)
def copy_target(target_path):
from shutil import copy
filename = os.path.basename(target_path)
src_path = os.path.join(core_src_dir, filename)
if not os.path.isfile(src_path):
src_path = os.path.join(editor_src_dir, filename)
if not os.path.isfile(src_path):
src_path = os.path.join(plugins_src_dir, filename)
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
print(f"Copying assembly to {target_path}...")
copy(src_path, target_path)
for scons_target in targets:
copy_target(scons_target)
return 0
def generate_sdk_package_versions():
# I can't believe importing files in Python is so convoluted when not
# following the golden standard for packages/modules.
import os
import sys
from os.path import dirname
# We want ../../../methods.py.
script_path = dirname(os.path.abspath(__file__))
root_path = dirname(dirname(dirname(script_path)))
sys.path.insert(0, root_path)
from methods import get_version_info
version_info = get_version_info("")
sys.path.remove(root_path)
version_str = "{major}.{minor}.{patch}".format(**version_info)
version_status = version_info["status"]
if version_status != "stable": # Pre-release
# If version was overridden to be e.g. "beta3", we insert a dot between
# "beta" and "3" to follow SemVer 2.0.
import re
match = re.search(r"[\d]+$", version_status)
if match:
pos = match.start()
version_status = version_status[:pos] + "." + version_status[pos:]
version_str += "-" + version_status
import version
version_defines = (
[
f"GODOT{version.major}",
f"GODOT{version.major}_{version.minor}",
f"GODOT{version.major}_{version.minor}_{version.patch}",
]
+ [f"GODOT{v}_OR_GREATER" for v in range(4, version.major + 1)]
+ [f"GODOT{version.major}_{v}_OR_GREATER" for v in range(0, version.minor + 1)]
+ [f"GODOT{version.major}_{version.minor}_{v}_OR_GREATER" for v in range(0, version.patch + 1)]
)
props = """<Project>
<PropertyGroup>
<PackageVersion_GodotSharp>{0}</PackageVersion_GodotSharp>
<PackageVersion_Godot_NET_Sdk>{0}</PackageVersion_Godot_NET_Sdk>
<PackageVersion_Godot_SourceGenerators>{0}</PackageVersion_Godot_SourceGenerators>
<GodotVersionConstants>{1}</GodotVersionConstants>
</PropertyGroup>
</Project>
""".format(version_str, ";".join(version_defines))
# We write in ../SdkPackageVersions.props.
with open(os.path.join(dirname(script_path), "SdkPackageVersions.props"), "w", encoding="utf-8", newline="\n") as f:
f.write(props)
# Also write the versioned docs URL to a constant for the Source Generators.
constants = """namespace Godot.SourceGenerators
{{
// TODO: This is currently disabled because of https://github.com/dotnet/roslyn/issues/52904
#pragma warning disable IDE0040 // Add accessibility modifiers.
partial class Common
{{
public const string VersionDocsUrl = "https://docs.godotengine.org/en/{docs_branch}";
}}
}}
""".format(**version_info)
generators_dir = os.path.join(
dirname(script_path),
"editor",
"Godot.NET.Sdk",
"Godot.SourceGenerators",
"Generated",
)
os.makedirs(generators_dir, exist_ok=True)
with open(os.path.join(generators_dir, "Common.Constants.cs"), "w", encoding="utf-8", newline="\n") as f:
f.write(constants)
def build_all(
msbuild_tool, module_dir, output_dir, godot_platform, dev_debug, push_nupkgs_local, precision, no_deprecated
):
# Generate SdkPackageVersions.props and VersionDocsUrl constant
generate_sdk_package_versions()
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
# Godot API
exit_code = build_godot_api(msbuild_tool, module_dir, output_dir, push_nupkgs_local, precision, no_deprecated)
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
if exit_code != 0:
return exit_code
# GodotTools
sln = os.path.join(module_dir, "editor/GodotTools/GodotTools.sln")
args = ["/restore", "/t:Build", "/p:Configuration=" + ("Debug" if dev_debug else "Release")] + (
["/p:GodotPlatform=" + godot_platform] if godot_platform else []
)
if push_nupkgs_local:
args += ["/p:ClearNuGetLocalCache=true", "/p:PushNuGetToLocalSource=" + push_nupkgs_local]
if precision == "double":
2022-09-02 17:08:40 +00:00
args += ["/p:GodotFloat64=true"]
exit_code = run_msbuild(msbuild_tool, sln=sln, chdir_to=module_dir, msbuild_args=args)
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
if exit_code != 0:
return exit_code
# Godot.NET.Sdk
args = ["/restore", "/t:Build", "/p:Configuration=Release"]
if push_nupkgs_local:
args += ["/p:ClearNuGetLocalCache=true", "/p:PushNuGetToLocalSource=" + push_nupkgs_local]
if precision == "double":
2022-09-02 17:08:40 +00:00
args += ["/p:GodotFloat64=true"]
if no_deprecated:
args += ["/p:GodotNoDeprecated=true"]
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
sln = os.path.join(module_dir, "editor/Godot.NET.Sdk/Godot.NET.Sdk.sln")
exit_code = run_msbuild(msbuild_tool, sln=sln, chdir_to=module_dir, msbuild_args=args)
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
if exit_code != 0:
return exit_code
return 0
def main():
import argparse
import sys
parser = argparse.ArgumentParser(description="Builds all Godot .NET solutions")
parser.add_argument("--godot-output-dir", type=str, required=True)
parser.add_argument(
"--dev-debug",
action="store_true",
default=False,
help="Build GodotTools and Godot.NET.Sdk with 'Configuration=Debug'",
)
parser.add_argument("--godot-platform", type=str, default="")
parser.add_argument("--mono-prefix", type=str, default="")
parser.add_argument("--push-nupkgs-local", type=str, default="")
parser.add_argument(
"--precision", type=str, default="single", choices=["single", "double"], help="Floating-point precision level"
)
parser.add_argument(
"--no-deprecated",
action="store_true",
default=False,
help="Build GodotSharp without using deprecated features. This is required, if the engine was built with 'deprecated=no'.",
)
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
args = parser.parse_args()
this_script_dir = os.path.dirname(os.path.realpath(__file__))
module_dir = os.path.abspath(os.path.join(this_script_dir, os.pardir))
output_dir = os.path.abspath(args.godot_output_dir)
push_nupkgs_local = os.path.abspath(args.push_nupkgs_local) if args.push_nupkgs_local else None
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
msbuild_tool = find_any_msbuild_tool(args.mono_prefix)
if msbuild_tool is None:
print("Unable to find MSBuild")
sys.exit(1)
exit_code = build_all(
msbuild_tool,
module_dir,
output_dir,
args.godot_platform,
args.dev_debug,
push_nupkgs_local,
args.precision,
args.no_deprecated,
)
C#: Move marshaling logic and generated glue to C# We will be progressively moving most code to C#. The plan is to only use Mono's embedding APIs to set things at launch. This will make it much easier to later support CoreCLR too which doesn't have rich embedding APIs. Additionally the code in C# is more maintainable and makes it easier to implement new features, e.g.: runtime codegen which we could use to avoid using reflection for marshaling everytime a field, property or method is accessed. SOME NOTES ON INTEROP We make the same assumptions as GDNative about the size of the Godot structures we use. We take it a bit further by also assuming the layout of fields in some cases, which is riskier but let's us squeeze out some performance by avoiding unnecessary managed to native calls. Code that deals with native structs is less safe than before as there's no RAII and copy constructors in C#. It's like using the GDNative C API directly. One has to take special care to free values they own. Perhaps we could use roslyn analyzers to check this, but I don't know any that uses attributes to determine what's owned or borrowed. As to why we maily use pointers for native structs instead of ref/out: - AFAIK (and confirmed with a benchmark) ref/out are pinned during P/Invoke calls and that has a cost. - Native struct fields can't be ref/out in the first place. - A `using` local can't be passed as ref/out, only `in`. Calling a method or property on an `in` value makes a silent copy, so we want to avoid `in`. REGARDING THE BUILD SYSTEM There's no longer a `mono_glue=yes/no` SCons options. We no longer need to build with `mono_glue=no`, generate the glue and then build again with `mono_glue=yes`. We build only once and generate the glue (which is in C# now). However, SCons no longer builds the C# projects for us. Instead one must run `build_assemblies.py`, e.g.: ```sh %godot_src_root%/modules/mono/build_scripts/build_assemblies.py \ --godot-output-dir=%godot_src_root%/bin \ --godot-target=release_debug` ``` We could turn this into a custom build target, but I don't know how to do that with SCons (it's possible with Meson). OTHER NOTES Most of the moved code doesn't follow the C# naming convention and still has the word Mono in the names despite no longer dealing with Mono's embedding APIs. This is just temporary while transitioning, to make it easier to understand what was moved where.
2021-05-03 13:21:06 +00:00
sys.exit(exit_code)
if __name__ == "__main__":
main()