chromium: Update to 138.0.7204.35

[caneraltinbasak]

Chromium 138+ uses its bundled libc++ and compiler-rt. To ensure the build system uses these and does not fall back to system-provided runtime libraries, explicitly pass the target triple and disable system stdlib++.

This avoids linker errors caused by missing crtbeginS.o, crtendS.o, and libgcc when using Clang with a minimal sysroot.

Release Notes:
https://chromereleases.googleblog.com/2025/07/stable-channel-update-for-desktop_15.html

Build and patch changes:
Added patches:

0011-fix-check_version-Only-compare-node.js-major-version.patch Updated patches:

0009-Adjust-the-Rust-build-to-our-needs.patch
Removed patches:

0011-Disable-crabbyavif-to-fix-build-errors.patch
0012-Revert-Allow-and-use-std-hardware_destructive_interf.patch 0013-Revert-Set-Rust-symbol-visibility-to-hidden-when-C-s.patch 0014-pdfium-Fix-missing-PDFiumAPIStringBufferAdapter-temp.patch 0015-Revert-Connect-the-Rust-log-crate-to-the-base-loggin.patch License changes:
Added licenses:

third_party/compiler-rt/src/LICENSE.TXT
third_party/libtess2/LICENSE
third_party/llvm-libc/src/LICENSE.TXT
third_party/net/third_party/mozilla_security_manager/LICENSE Removed licenses:

third_party/accessibility-audit/LICENSE
third_party/io_grpc_grpc_api/LICENSE
third_party/libavifinfo/LICENSE
Updated licenses:

third_party/android_deps/libs/com_google_android_gms_play_services_auth/LICENSE → third_party/android_deps/autorolled/committed/libs/com_google_android_gms_play_services_auth/LICENSE third_party/android_deps/libs/com_google_android_gms_play_services_auth_api_phone/LICENSE → third_party/android_deps/autorolled/committed/libs/com_google_android_gms_play_services_auth_api_phone/LICENSE third_party/android_deps/libs/com_google_android_gms_play_services_auth_base/LICENSE → third_party/android_deps/autorolled/committed/libs/com_google_android_gms_play_services_auth_base/LICENSE third_party/android_deps/libs/com_google_android_gms_play_services_auth_blockstore/LICENSE → third_party/android_deps/autorolled/committed/libs/com_google_android_gms_play_services_auth_blockstore/LICENSE third_party/android_deps/libs/com_google_android_gms_play_services_fido/LICENSE → third_party/android_deps/autorolled/committed/libs/com_google_android_gms_play_services_fido/LICENSE third_party/android_deps/libs/com_google_android_libraries_identity_googleid_googleid/LICENSE → third_party/android_deps/autorolled/committed/libs/com_google_android_libraries_identity_googleid_googleid/LICENSE third_party/expat/src/expat/COPYING
third_party/libjpeg_turbo/LICENSE
Test-built
chromium-ozone-wayland

master, clang, MACHINE=qemux86-64
chromium-x11

master, clang, MACHINE=qemux86-64, qemuarm, qemuarm64 Repo hashes (from meta-chromium-test)
meta-oe:
commit: "a8dfd10600035a799abae03178fc7054582ea43d" poky:
commit: "243d54fd466f5f852cc0fdcce57997918ce35f32" meta-clang:
commit: "7a2f83360920b10214e2659e17a4b9cba2d0435b"

[caneraltinbasak]

Verified for x11, ozone for arm, aarch64 and x86-64:
https://github.com/brightsign/meta-browser/actions/runs/17671009313

[zboszor]

Build error from today, after oe-core switched to CLANG 21.1.0:

Log data follows:
| DEBUG: Executing shell function do_copy_clang_library
| cp: cannot create regular file '/mnt/zozo/yocto-5.3-arm/tmp-sicom/work/corei7-64-oe-linux/chromium-x11/138.0.7204.157/recipe-sysroot-native/usr/lib/clang/latest/lib/linux/libclang_rt.builtins-x86_64.a': No such file or directory
| WARNING: /mnt/zozo/yocto-5.3-arm/tmp-sicom/work/corei7-64-oe-linux/chromium-x11/138.0.7204.157/temp/run.do_copy_clang_library.1112517:157 exit 1 from 'cp "$lib_file" "/mnt/zozo/yocto-5.3-arm/tmp-sicom/work/corei7-64-oe-linux/chromium-x11/138.0.7204.157/recipe-sysroot-native/usr/lib/clang/latest/${stripped_lib_file}"'
| WARNING: Backtrace (BB generated script):
| 	#1: do_copy_clang_library, /mnt/zozo/yocto-5.3-arm/tmp-sicom/work/corei7-64-oe-linux/chromium-x11/138.0.7204.157/temp/run.do_copy_clang_library.1112517, line 157
| 	#2: main, /mnt/zozo/yocto-5.3-arm/tmp-sicom/work/corei7-64-oe-linux/chromium-x11/138.0.7204.157/temp/run.do_copy_clang_library.1112517, line 160

Please, fix it. FWIW, I use meta-intel and MACHINE="intel-corei7-64".

[caneraltinbasak]

Can you please try your target machine compile with following hashes checked out:

meta-oe: "a8dfd10600035a799abae03178fc7054582ea43d"
poky: "243d54fd466f5f852cc0fdcce57997918ce35f32"
meta-clang: "7a2f83360920b10214e2659e17a4b9cba2d0435b"

I've used the hashes Firefox was using for testing Whinlatter branch, they are 3 months old. I would like to land my changes as they are and do adaptations for the latest yocto release as a follow up change.

[OldManYellsAtCloud]

I think until than it should be kept as Walnascar-compatible only maybe. (If I remember well, those hashes are essentially the last Walnascar-compatible revisions after the actual Walnascar release, just before some breaking changes were introduced. FF has switched to the tip of master since then, though FF is a much smaller beast than Chromium)

[caneraltinbasak]

I think until than it should be kept as Walnascar-compatible only maybe. (If I remember well, those hashes are essentially the last Walnascar-compatible revisions after the actual Walnascar release, just before some breaking changes were introduced. FF has switched to the tip of master since then, though FF is a much smaller beast than Chromium)

That is fine. I can target my changes to walnascar branch, if a maintainer creates a walnascar branch.

[caneraltinbasak]

meta-oe: "a8dfd10600035a799abae03178fc7054582ea43d" poky: "243d54fd466f5f852cc0fdcce57997918ce35f32" meta-clang: "7a2f83360920b10214e2659e17a4b9cba2d0435b"

I've used the hashes Firefox was using for testing Whinlatter branch, they are 3 months old. I would like to land my changes as they are and do adaptations for the latest yocto release as a follow up change.

Sorry, I have to correct myself here. I checked again where these hashes are coming from. It is not Firefox, these are tip of meta-oe, poky and meta-clang as of 1st of July when I first started upgrading meta-chromium to Chromium138 and poky had layer name as whinlatter at the time.

I had vacation and other work interruptions in between...

[zboszor]

meta-oe: "a8dfd10600035a799abae03178fc7054582ea43d" poky: "243d54fd466f5f852cc0fdcce57997918ce35f32" meta-clang: "7a2f83360920b10214e2659e17a4b9cba2d0435b"
I've used the hashes Firefox was using for testing Whinlatter branch, they are 3 months old. I would like to land my changes as they are and do adaptations for the latest yocto release as a follow up change.

Sorry, I have to correct myself here. I checked again where these hashes are coming from. It is not Firefox, these are tip of meta-oe, poky and meta-clang as of 1st of July when I first started upgrading meta-chromium to Chromium138 and poky had layer name as whinlatter at the time.

I had no problems at around the beginning of summer either, but since then a lot of breaking changes were added to oe-core for clang and related recipes.

[adalessandro]

@caneraltinbasak pushed fixes on brightsign#15 , please review and merge.