Distcc somehow no longer works with clang

[motig]

@vr13 I'm also using your patch to get chromium to compile, thank you for "maintaining" it. I am actually using icecream and the behaviour is the same as distcc, i.e. I have to use this patch lest the build fail.

[vr13]

I am actually using icecream and the behaviour is the same as distcc, i.e. I have to use this patch lest the build fail.

thank you for the info. Will keep staying with distcc then

[mortonP]

Patch still works with chromium-142.0.7444.134 :-)

[mortonP]

trying with chromium 143.0.7499.40...

patch seems to still apply -- however doesn't distcc build because of:

worker side:
distccd[4741] (dcc_job_summary) client: 192.168.1.89:36914 COMPILE_ERROR exit:1 sig:0 core:0 ret:0 time:167ms x86_64-pc-linux-gnu-clang++-20 ../../third_party/abseil-cpp/absl/base/internal/unscaledcycleclock.cc
distccd[4739] (dcc_job_summary) client: 192.168.1.89:36870 COMPILE_ERROR exit:1 sig:0 core:0 ret:0 time:175ms x86_64-pc-linux-gnu-clang++-20 ../../third_party/abseil-cpp/absl/base/internal/cycleclock.cc
....
istccd[4727] (dcc_job_summary) client: 192.168.1.89:54812 COMPILE_ERROR exit:1 sig:0 core:0 ret:0 time:625ms x86_64-pc-linux-gnu-clang++-20 ../../third_party/protobuf/src/google/protobuf/any.cc
distccd[4731] (dcc_job_summary) client: 192.168.1.89:54822 COMPILE_ERROR exit:1 sig:0 core:0 ret:0 time:612ms x86_64-pc-linux-gnu-clang++-20 ../../third_party/protobuf/src/google/protobuf/compiler/importer.cc
distccd[4733] (dcc_job_summary) client: 192.168.1.89:54832 COMPILE_ERROR exit:1 sig:0 core:0 ret:0 time:704ms x86_64-pc-linux-gnu-clang++-20 ../../third_party/protobuf/src/google/protobuf/compiler/parser.cc
....
distccd[4730] (dcc_job_summary) client: 192.168.1.89:52550 COMPILE_ERROR exit:1 sig:0 core:0 ret:0 time:201ms x86_64-pc-linux-gnu-clang++-20 ../../net/third_party/quiche/src/quiche/common/vectorized_io_utils.cc
distccd[4740] (dcc_job_summary) client: 192.168.1.89:33300 COMPILE_ERROR exit:1 sig:0 core:0 ret:0 time:464ms x86_64-pc-linux-gnu-clang++-20 ../../net/third_party/quiche/src/quiche/quic/core/http/spdy_utils.cc
...

client side:
x86_64-pc-linux-gnu-clang++-20: error: no such file or directory: '../../tools/ubsan/array_bounds_ignorelist.txt'
x86_64-pc-linux-gnu-clang++-20: error: no such file or directory: '../../tools/ubsan/array_bounds_ignorelist.txt'
x86_64-pc-linux-gnu-clang++-20: error: no such file or directory: '../../tools/ubsan/array_bounds_ignorelist.txt'
x86_64-pc-linux-gnu-clang++-20: error: no such file or directory: '../../tools/ubsan/array_bounds_ignorelist.txt'

...any idea what's going on here with 143?

[vr13]

client side:
x86_64-pc-linux-gnu-clang++-20: error: no such file or directory: '../../tools/ubsan/array_bounds_ignorelist.txt'

Try this. One contains previous changes plus your recent finding:

Code: Select all

--- a/build/config/compiler/BUILD.gn    2025-09-13 17:15:05.585891128 +0300
+++ b/build/config/compiler/BUILD.gn    2025-09-13 17:10:57.398681527 +0300

@@ -1707,10 +1707,7 @@
 config("clang_warning_suppression") {
   # Some build configs use older versions of clang that don't support WSMs
   if (use_clang_warning_suppression_file) {
-    from_build_root =
-        rebase_path(clang_warning_suppression_file, root_build_dir)
-    inputs = [ clang_warning_suppression_file ]
-    cflags = [ "--warning-suppression-mappings=" + from_build_root ]
+    cflags = [ "-w" ]
   }
 }
 
@@ -1735,10 +1732,6 @@
       "-fsanitize-trap=array-bounds",
     ]
 
-    if (defined(ubsan_array_bounds_ignorelist_path)) {
-      rebased = rebase_path(ubsan_array_bounds_ignorelist_path, root_build_dir)
-      cflags += [ "-fsanitize-ignorelist=$rebased" ]
-    }
   }
 }

--- a/build/linux/unbundle/openh264.gn  2025-09-09 02:21:33.000000000 +0300
+++ b/build/linux/unbundle/openh264.gn  2025-09-13 17:11:37.951153925 +0300
@@ -6,13 +6,7 @@
 import("//build/shim_headers.gni")
 
 config("encoder_cfi_flags") {
-  cflags = []
-  if (current_toolchain == default_toolchain && is_clang) {
-    openh264_cfi_ignorelist_path =
-        rebase_path("//build/linux/unbundle/openh264_encoder_cfi_ignores.txt",
-                    root_build_dir)
-    cflags += [ "-fsanitize-ignorelist=$openh264_cfi_ignorelist_path" ]
-  }
+  cflags = [ "-w" ]
 }
 
 pkg_config("config") {

thanks

[mortonP]

Try this.

Went through - from a quick browse I only see "COMPILE_OK exit:0" in /var/log/distccd.log and runtime also feels approximately right.

Thank you for helping keeping my old laptop from being decommissioned! :-)
(and a *grrrrr* to websites that require a Chrome-engine browser to work properly...)

[mortonP]

patch fails with chromium-144.0.7559.109:

Code: Select all

* Applying 02-distcc.patch ...
patching file build/config/compiler/BUILD.gn
Hunk #1 succeeded at 1701 (offset -6 lines).
Hunk #2 FAILED at 1732.
1 out of 2 hunks FAILED -- saving rejects to file build/config/compiler/BUILD.gn.rej
patching file build/linux/unbundle/openh264.gn
[ !! ]
 * ERROR: www-client/chromium-144.0.7559.109::gentoo failed (prepare phase):

...btw, how to bribe someone to add this patch to chromium ebuilds by default?

[vr13]

patch fails with chromium-144.0.7559.109:

Code: Select all

--- a/build/config/compiler/BUILD.gn       2026-02-06 09:03:18.496535389 +0300
+++ b/build/config/compiler/BUILD.gn       2026-02-06 09:25:26.124644545 +0300
@@ -1701,10 +1701,7 @@
 config("clang_warning_suppression") {
   # Some build configs use older versions of clang that don't support WSMs
   if (use_clang_warning_suppression_file) {
-    from_build_root =
-        rebase_path(clang_warning_suppression_file, root_build_dir)
-    inputs = [ clang_warning_suppression_file ]
-    cflags = [ "--warning-suppression-mappings=" + from_build_root ]
+    cflags = [ "-w" ]
   }
 }

--- a/build/linux/unbundle/openh264.gn  2025-09-09 02:21:33.000000000 +0300
+++ b/build/linux/unbundle/openh264.gn  2025-09-13 17:11:37.951153925 +0300
@@ -6,13 +6,7 @@
 import("//build/shim_headers.gni")
 
 config("encoder_cfi_flags") {
-  cflags = []
-  if (current_toolchain == default_toolchain && is_clang) {
-    openh264_cfi_ignorelist_path =
-        rebase_path("//build/linux/unbundle/openh264_encoder_cfi_ignores.txt",
-                    root_build_dir)
-    cflags += [ "-fsanitize-ignorelist=$openh264_cfi_ignorelist_path" ]
-  }
+  cflags = [ "-w" ]
 }
 
 pkg_config("config") {

[vr13]

...btw, how to bribe someone to add this patch to chromium ebuilds by default?

There are two choices: either file a bug or create a user patch :)

The idea behind that patch is almost trivial:

• starting from version 137 chromium build system employs functionality to selectively filter clang warning [types/classes whatever]. These are listed in provided .txt files
• distcc (and sys-devel/icecream -- confirmed by @motig few posts above) does not support that filtering and stumbles on attempt to compile a source with `external file filter rule'
• I simply excluded all `external file filter rules' from build/config/compiler/BUILD.gn
• additionally I updated that BUILD.gn template with "-w" to suppress all compiler warnings thus solving additional issue reported here

I guess it should be more efficient way to do that. The issue could be fixed with either improved distcc or just in chromium ebuild patch, I do now know what is better and choose simplest approach (portage user patch, shared here)

[mortonP]

file a bug

There is already a bug about chromium and distcc https://bugs.gentoo.org/870085 ...which has stopped in 2024.

could be fixed with either improved distcc or just in chromium ebuild patch

Well, a patch carried with chromium would at least be an improvement, before e.g. distcc is improved.

For me, without distcc, chromium is a ~1.5 day compile...

[vr13]

file a bug

There is already a bug about chromium and distcc https://bugs.gentoo.org/870085 ...which has stopped in 2024.

thank you. I did not notice it when started playing with chromium ebuild last summer. Actually the bug is not only about the issue we're talking in here. For me the proposal to have kinda distcc wrapper looks a bit ugly. The problem however was touched

I agree the problem is not so significant and is relevant only to those who still compiles chromium from source and do not want to use binary package (like me :)

For me, without distcc, chromium is a ~1.5 day compile...

just finished with 144 ebuild in 4 hours..

[mortonP]

just finished with 144 ebuild in 4 hours..

On my old laptop with two distcc workers I get down from ~1.5 days to ~5.5h

On a newer laptop I get now ~1h50m. Guess with wired instead of Wifi this would even be a bit faster.

Again, thank you for the patch!

[alexbuell]

thanks for this, this has been bugging me for quite a while. I'll give it a go with my upcoming Chromium 144 build.

[vr13]

FYI: recent patch also works with chromium 145

[vr13]

seems the patch prepared for 144 build still works with 146. At least overnight build just finished:

Code: Select all

$ qlop -v --date yesterday
...
2026-03-14T23:34:39 >>> www-client/chromium-146.0.7680.80: 4:38:39