yams/plugins/zyp/meson.build
trvon 93f2439471
build(plugins): export only the plugin C ABI; hide onnx_resource internals
Plugins and libyams_onnx_resource were linked with default visibility, so
their dynamic symbol tables exported everything they contained, including
statically linked third-party code. The S3 plugin exported 11,147
symbols (OpenSSL, libcurl, zlib, spdlog, nlohmann_json); the ONNX
resource library exported 895 (850 of them spdlog). Every loader uses
RTLD_LOCAL, so this was not a live interposition bug, but the export
table is the ABI and anything that binds RTLD_GLOBAL (or a future host)
could resolve into a plugin's private copies.

Plugins:
- plugins/exports/yams_plugin.map (ELF version script, global
  yams_plugin_*, local *) and yams_plugin.exp (ld64
  -exported_symbols_list) are applied to every plugin (s3, onnx, glint,
  zyp) with gnu_symbol_visibility: 'hidden' and --exclude-libs,ALL.
- The S3 legacy factory (yams_plugin_create/destroy_object_storage) is
  marked YAMS_PLUGIN_API so it stays exported under hidden visibility.

libyams_onnx_resource:
- Builds with hidden visibility; YAMS_ONNX_RESOURCE_API now expands to
  visibility("default") on GCC/Clang, so only OnnxConcurrencyRegistry is
  exported.
- It no longer links spdlog. Its messages reached the host log only
  because the executable's spdlog interposed the library's copy; with the
  copy hidden, the registry would have logged through a private default
  logger to stdout. The registry now takes a host log sink
  (OnnxConcurrencyRegistry::setLogSink); TuningManager and
  AbiModelProviderAdapter install a spdlog forwarder from the header-only
  onnx_registry_spdlog_sink.h. Without a sink, messages are dropped.

Tests (red first on grim, Debug):
- tests/plugins/plugin_export_surface_catch2_test.cpp dlopens the built
  S3 and zyp plugins and asserts the entry points resolve while libcurl,
  OpenSSL, zlib and zyp internals do not: 2/2 cases failed (11 of 31
  assertions) before, pass after.
- resource_governance_test gains a case for the log sink routing (did
  not compile before; passes after). `meson test --suite plugins`: 5/5.

Measured on grim (release, stripped staged payload):

  file                          bytes before  after       exports before  after
  libyams_object_storage_s3.so   7,854,064   6,240,112          11,147     10
  yams_zyp.so                       80,112      63,728             103      8
  libyams_onnx_resource.so         477,472      30,840             895     20
  bin/yams-cli                  33,712,056  33,859,512              80     39
  bin/yams-daemon               28,389,784  28,537,304              98     39

The executables now carry their own spdlog code paths instead of
exporting them for the resource library to interpose (+0.4%), and export
41 fewer symbols each. check_install_surface turns GREEN; budgets are
reset to the measured payload plus ~20%.

Signed-off-by: trvon <me@trevon.dev>
2026-10-01 18:51:34 -06:00

222 lines
7.1 KiB
Meson

#zyp - Zig YAMS PDF Extractor
#High - performance PDF text extraction using zpdf(https: // github.com/Lulzx/zpdf)
if not get_option ('plugin-zyp')
subdir_done()
endif
if host_machine.system() == 'darwin' and get_option('b_coverage')
warning('Skipping zyp plugin in macOS coverage builds: Clang gcov runtime cannot link with the zyp bundle/benchmark.')
subdir_done()
endif
#Find Zig compiler(requires 0.15.x; zpdf is not Zig 0.16 - compatible yet)
zig = find_program('zig', required: false)
if not zig.found()
warning('Zig compiler not found, skipping zyp plugin. Install Zig 0.15.x to enable.')
subdir_done()
endif
zig_version_result = run_command(zig, 'version', check: false)
if zig_version_result.returncode() != 0
warning('Unable to determine Zig version, skipping zyp plugin.')
subdir_done()
endif
zig_version = zig_version_result.stdout().strip()
if not zig_version.version_compare('>=0.15.0')
warning('Zig @0@ is unsupported by zyp; zpdf requires Zig 0.15+. Skipping zyp plugin.'.format(zig_version))
subdir_done()
endif
#Platform - specific library name and rpath
if host_machine.system() == 'darwin'
zpdf_lib_name = 'libzpdf.dylib'
#@loader_path is used for shared libraries / modules to find libs in same directory
zyp_install_rpath = '@loader_path'
zyp_link_args = ['-fno-sanitize=thread']
elif host_machine.system() == 'windows'
zpdf_lib_name = 'zpdf.dll'
zpdf_import_lib_name = 'zpdf.lib'
zyp_install_rpath = '' # Windows doesn't use rpath
#Link against dynamic CRT libraries; avoid static CRT conflict
zyp_link_args = [
'/NODEFAULTLIB:libcmt',
'/NODEFAULTLIB:libcmtd',
'msvcrt.lib',
'vcruntime.lib',
'ucrt.lib',
]
else
zpdf_lib_name = 'libzpdf.so'
#$ORIGIN refers to directory containing the loading module
zyp_install_rpath = '$ORIGIN'
zyp_link_args = ['-fno-sanitize=thread']
endif
#Build zpdf shared library via Zig and copy to build directory
#Zig outputs DLLs to zpdf / zig - out / bin / on Windows, shared libs to zig - out / lib / on Unix
zpdf_src_dir = meson.current_source_dir() / 'zpdf'
if host_machine.system() == 'windows'
#Windows : DLL goes to bin /, import lib goes to lib /
zpdf_lib_src = zpdf_src_dir / 'zig-out' / 'bin' / zpdf_lib_name
zpdf_import_lib_src = zpdf_src_dir / 'zig-out' / 'lib' / zpdf_import_lib_name
else
zpdf_lib_src = zpdf_src_dir / 'zig-out' / 'lib' / zpdf_lib_name
endif
zpdf_build_dir = meson.current_build_dir()
fs = import('fs')
if not fs.exists(zpdf_src_dir / 'build.zig')
warning('zpdf source not found (missing build.zig), skipping zyp plugin build.')
subdir_done()
endif
#Platform - specific build command
if host_machine.system() == 'windows'
#Windows : Use PowerShell for build
#Build zpdf and copy both DLL and import lib to build directory
zpdf_lib = custom_target('zpdf_shared',
command: [
'powershell', '-NoProfile', '-Command',
'Push-Location "@0@"; & "@1@" build -Doptimize=ReleaseFast; Copy-Item "@2@" "@3@\\@4@"; Copy-Item "@5@" "@3@\\@6@"; Pop-Location'.format(
zpdf_src_dir,
zig.full_path(),
zpdf_lib_src,
zpdf_build_dir,
zpdf_lib_name,
zpdf_import_lib_src,
zpdf_import_lib_name
)
],
output: [zpdf_lib_name, zpdf_import_lib_name],
build_by_default: true,
console: true,
)
#Extract just the import library for linking
zpdf_import_lib = zpdf_lib[1]
elif host_machine.system() == 'darwin'
zpdf_build_command = (
'cd "@0@" && "@1@" build -Doptimize=ReleaseFast -Dtarget=@2@-macos.${MACOSX_DEPLOYMENT_TARGET:-26.0} && cp "@3@" "@4@/@5@"'
).format(
zpdf_src_dir,
zig.full_path(),
host_machine.cpu_family(),
zpdf_lib_src,
zpdf_build_dir,
zpdf_lib_name
)
zpdf_lib = custom_target('zpdf_shared',
command: [
'sh', '-c', zpdf_build_command
],
output: [zpdf_lib_name],
build_by_default: true,
console: true,
)
else
#Unix(Linux / macOS) : Use shell
zpdf_lib = custom_target('zpdf_shared',
command: [
'sh', '-c',
'cd "@0@" && "@1@" build -Doptimize=ReleaseFast && cp "@2@" "@3@/@4@"'.format(
zpdf_src_dir,
zig.full_path(),
zpdf_lib_src,
zpdf_build_dir,
zpdf_lib_name
)
],
output: [zpdf_lib_name],
build_by_default: true,
console: true,
)
endif
#Create a dependency object that includes the exact custom-target output. Meson recognizes
#the shared/import-library suffix and passes this artifact to the linker while preserving its
#generated-file edge. Avoid a separate -L/-l lookup: Meson resolves that to an absolute input
#without a generating rule, which breaks a clean rebuild after the artifact is removed.
if host_machine.system() == 'windows'
#On Windows, link against the import library only (not the DLL).
zpdf_dep = declare_dependency(sources: [zpdf_import_lib])
else
zpdf_dep = declare_dependency(sources: [zpdf_lib])
endif
#Compiler - specific flags for this plugin.
zyp_cpp_args = []
if meson.get_compiler('cpp').get_id() != 'msvc'
zyp_cpp_args = ['-std=c++20', '-fno-sanitize=thread']
endif
#Plugin sources
zyp_sources = files(
'zyp_plugin.cpp',
'zpdf_wrapper.cpp',
'pdf_metadata.cpp',
)
#Build the plugin(minimal dependencies - no spdlog to avoid TSan conflicts)
zyp_plugin = shared_module('yams_zyp',
zyp_sources,
include_directories: [
include_directories('../../include'),
include_directories('.'),
],
dependencies: [zpdf_dep],
cpp_args: zyp_cpp_args,
link_args: zyp_link_args + plugin_export_link_args,
link_depends: plugin_export_link_depends,
gnu_symbol_visibility: 'hidden',
#build_rpath ensures the plugin can find libzpdf during build / test
build_rpath: zpdf_build_dir,
#install_rpath ensures the plugin can find libzpdf after installation
install_rpath: zyp_install_rpath,
build_by_default: true,
install: true,
install_dir: plugins_dir,
name_prefix: '',
)
#Install zpdf library alongside the plugin
#Use platform - specific install script
if host_machine.system() == 'windows'
meson.add_install_script('powershell', '-NoProfile', '-Command',
'$dest = if ($env:DESTDIR) { Join-Path $env:DESTDIR "@1@" } else { "@1@" }; New-Item -ItemType Directory -Force -Path $dest | Out-Null; Copy-Item -Path "@0@" -Destination $dest'.format(
zpdf_build_dir / zpdf_lib_name,
get_option('prefix') / plugins_dir
),
install_tag: 'runtime',
)
else
meson.add_install_script('sh', '-c',
'mkdir -p "$DESTDIR@1@" && cp "@0@" "$DESTDIR@1@/"'.format(
zpdf_build_dir / zpdf_lib_name,
get_option('prefix') / plugins_dir
),
install_tag: 'runtime',
)
endif
#-- -- zyp extraction benchmark -- --
#Measures sequential vs parallel PDF extraction throughput.
#Usage : zyp_extraction_bench < pdf_path>[repeat_count]
zyp_extraction_bench_sources = files(
'../../tests/benchmarks/zyp_extraction_bench.cpp',
'zpdf_wrapper.cpp',
)
zyp_extraction_bench = executable('zyp_extraction_bench',
zyp_extraction_bench_sources,
include_directories: [
include_directories('../../include'),
include_directories('../..'),
include_directories('.'),
],
dependencies: [zpdf_dep],
cpp_args: zyp_cpp_args,
build_rpath: zpdf_build_dir,
build_by_default: false,
install: false,
override_options: ['b_sanitize=none'],
)