| Summary: | www-client/firefox-112.0[pgo]: fails to install (error: $SRCDIR/browser/installer/package-manifest.in:48: Missing file(s)) | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Michael Mair-Keimberger (mm1ke) <mmk> |
| Component: | Current packages | Assignee: | Mozilla Gentoo Team <mozilla> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | alex_y_xu |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| See Also: | https://bugs.gentoo.org/show_bug.cgi?id=903710 | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: | firefox-112.build.log | ||
|
Description
Michael Mair-Keimberger (mm1ke)
2023-04-15 07:17:34 UTC
it might be unhappy about missing L10N (In reply to tt_1 from comment #1) > it might be unhappy about missing L10N Thanks for the hint. I've added en-GB now but unfortunately it still fails the same. [34m 7:51.03(B[m gmake[1]: Leaving directory '/var/tmp/portage/www-client/firefox-112.0/work/firefox_build/instrumented/browser/installer'(B[m(B[m
[34m 7:51.03(B[m gmake: Leaving directory '/var/tmp/portage/www-client/firefox-112.0/work/firefox_build/instrumented'(B[m(B[m
00:07:51.613 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac5299cae0 (res 0x55ac5299ce40)[0m
00:07:51.613 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac5299cf60 (res 0x55ac5299d2c0)[0m
console.warn: services.settings: Ignoring preference override of remote settings server
console.warn: services.settings: Allow by setting MOZ_REMOTE_SETTINGS_DEVTOOLS=1 in the environment
00:07:51.858 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac5299d590 (res 0x55ac5299d8f0)[0m
00:07:51.858 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac5299d8f0 (res 0x55ac5299dc50)[0m
00:07:51.925 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac5299e1c0 (res 0x55ac5299e520)[0m
00:07:52.089 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac5299f650 (res 0x55ac5299f9b0)[0m
00:07:52.089 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac529a5660 (res 0x55ac5299fa40)[0m
00:07:52.123 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac529a59c0 (res 0x55ac5299fad0)[0m
00:07:52.123 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac529a5d20 (res 0x55ac5299e520)[0m
00:07:52.132 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac529aad10 (res 0x55ac529ab070)[0m
00:07:52.132 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac529ab190 (res 0x55ac529ab4f0)[0m
00:07:52.148 [1;90m[types/wlr_compositor.c:673] New wlr_surface 0x55ac529ab580 (res 0x55ac529ab8e0)[0m
Exiting due to channel error.
Exiting due to channel error.
Firefox exited with code -11 during profile initialization
Error running mach:
['build', '--verbose']
The error occurred in code that was called by the mach command. This is either
a bug in the called code itself or in the way that mach is calling it.
You can invoke ``./mach busted`` to check if this issue is already on file. If it
isn't, please use ``./mach busted file build`` to report it. If ``./mach busted`` is
misbehaving, you can also inspect the dependencies of bug 1543241.
If filing a bug, please include the full output of mach, including this error
message.
The details of the failure are as follows:
subprocess.CalledProcessError: Command '['/var/tmp/portage/www-client/firefox-112.0/work/firefox_build/_virtualenvs/build/bin/python', '/var/tmp/portage/www-client/firefox-112.0/work/firefox-112.0/build/pgo/profileserver.py']' returned non-zero exit status 245.
File "/var/tmp/portage/www-client/firefox-112.0/work/firefox-112.0/python/mozbuild/mozbuild/build_commands.py", line 215, in build
subprocess.check_call(pgo_cmd, cwd=instr.topobjdir, env=pgo_env)
File "/usr/lib/python3.11/subprocess.py", line 413, in check_call
raise CalledProcessError(retcode, cmd)
>>> Source compiled.
Thanks sam for the update. I've checked the other bug. It seems to be the same problem, but unfortunately the problem still persist for me. For the record. I'm using gui-wm/sway here as my window manager, which get started via gui-apps/tuigreet. It starts a script which looks like this: #!/bin/sh export MOZ_ENABLE_WAYLAND=1 exec dbus-run-session sway >> /dev/null 2>&1 I'm wondering if i might miss some env variables. I've checked with the settings from the other bug and tried to set following (as they were empty for me): export XDG_CURRENT_DESKTOP=sway export XDG_SESSION_DESKTOP=sway export SDL_VIDEODRIVER=wayland So far without any luck. I know this sounds dumb, but could you reboot and immediately try compiling firefox again? I can't see any other reason except the graphical session somehow not interacting with the pgo build process. Although it could be the virtwl() function, it could also be sway. I guess I'll need a wayland-only VM to test this. https://bugzilla.mozilla.org/show_bug.cgi?id=1551279 maybe something in here proves helpful? Maybe https://hg.mozilla.org/mozilla-central/rev/1068e0955cfb does something? You can wget it from https://hg.mozilla.org/mozilla-central/raw-rev/1068e0955cfb and place under /etc/portage/patches. You could also try to remove the --enable-system-pixman line from our src_configure to see if that helps? ^ I'll include that patch for 112.0.1, doing test runs now. Will push in few hours if everything goes okay. The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=2c053bbe89752d9418b27f28eeb0e22eaa946db0 commit 2c053bbe89752d9418b27f28eeb0e22eaa946db0 Author: Joonas Niilola <juippis@gentoo.org> AuthorDate: 2023-04-18 08:35:12 +0000 Commit: Joonas Niilola <juippis@gentoo.org> CommitDate: 2023-04-18 08:35:12 +0000 www-client/firefox: add 112.0.1 Bug: https://bugs.gentoo.org/904354 Signed-off-by: Joonas Niilola <juippis@gentoo.org> www-client/firefox/Manifest | 101 +++ www-client/firefox/firefox-112.0.1.ebuild | 1379 +++++++++++++++++++++++++++++ 2 files changed, 1480 insertions(+) Hi Joonas,
thanks for the update. Unfortunately no changes for me yet. I even rebooted the system again to have a freshly booted system.
The error still looks pretty much the same:
00:08:01.167 [types/wlr_compositor.c:673] New wlr_surface 0x5626568cea90 (res 0x5626568cedf0)
00:08:01.167 [types/wlr_compositor.c:673] New wlr_surface 0x5626568cef10 (res 0x5626568cf270)
console.warn: services.settings: Ignoring preference override of remote settings server
console.warn: services.settings: Allow by setting MOZ_REMOTE_SETTINGS_DEVTOOLS=1 in the environment
00:08:01.406 [types/wlr_compositor.c:673] New wlr_surface 0x5626568cf540 (res 0x5626568cf8a0)
00:08:01.406 [types/wlr_compositor.c:673] New wlr_surface 0x5626568cf8a0 (res 0x5626568cfc00)
00:08:01.472 [types/wlr_compositor.c:673] New wlr_surface 0x5626568d00e0 (res 0x5626568d0440)
00:08:01.624 [types/wlr_compositor.c:673] New wlr_surface 0x5626568d1400 (res 0x5626568d1760)
00:08:01.624 [types/wlr_compositor.c:673] New wlr_surface 0x5626568d1880 (res 0x5626568d1be0)
00:08:01.653 [types/wlr_compositor.c:673] New wlr_surface 0x5626568d76b0 (res 0x5626568d07e0)
00:08:01.653 [types/wlr_compositor.c:673] New wlr_surface 0x5626568d7a10 (res 0x5626568d0440)
00:08:01.662 [types/wlr_compositor.c:673] New wlr_surface 0x5626568dcc40 (res 0x5626568dcfa0)
00:08:01.662 [types/wlr_compositor.c:673] New wlr_surface 0x5626568dd0c0 (res 0x5626568dd420)
00:08:01.667 [types/wlr_compositor.c:673] New wlr_surface 0x5626568dd4b0 (res 0x5626568dd810)
Exiting due to channel error.
Exiting due to channel error.
Firefox exited with code -11 during profile initialization
Error running mach:
['build', '--verbose']
The error occurred in code that was called by the mach command. This is either
a bug in the called code itself or in the way that mach is calling it.
You can invoke ``./mach busted`` to check if this issue is already on file. If it
isn't, please use ``./mach busted file build`` to report it. If ``./mach busted`` is
misbehaving, you can also inspect the dependencies of bug 1543241.
If filing a bug, please include the full output of mach, including this error
message.
The details of the failure are as follows:
subprocess.CalledProcessError: Command '['/var/tmp/portage/www-client/firefox-112.0.1/work/firefox_build/_virtualenvs/build/bin/python', '/var/tmp/portage/www-client/firefox-112.0.1/work/firefox-112.0.1/build/pgo/profileserver.py']' returned non-zero exit status 245
FYI: Forgot to test to build firefox with --enable-system-pixman removed from the ebuild. However, still the same issue. Some more information: I've now tried to rebuild firefox-111.0. (restored the ebuild and put it into my overlay). Firefox 111.0 fails now too. So maybe this is more related to some dependencies? Right, anything in dmesg? Has your graphical drivers crashed for example (although shouldn't happen after a fresh reboot?) Other than that really no clue. You should report it upstream, build.log all the way to that crash is relevant (your post in #c9). The src_install failure is irrelevant because that's an error in the ebuild. Please also link the upstream bug so when resolved we can resolve that in Gentoo too. Thanks Joonas for the help. I've checked dmesg for some errors. The only error i could find which could be related was: [ 51.311040] gldriverquery[6852]: segfault at 0 ip 0000000000000000 sp 00007ffc180d9fd8 error 14 in gldriverquery[560b73800000+14000] likely on CPU 31 (core 15, socket 0) This comes basically right after logging. However as far as I can see it doesn't happen every boot and last time i tried to compile firefox it didn't happen. I've also haven't seen any other problems because of that.. I can make an upstream bug, just to be sure, with upstream you mean firefox upstream? So, there's two problems here:
1. instrumented firefox crashes in Servo_AuthorStyles_Flush trying to write to libxul.so .rodata. I'm still investigating this issue.
2. virtwl has a bug, it doesn't propagate the return code. for this issue, please apply this patch:
diff --git a/www-client/firefox/firefox-112.0.1.ebuild b/www-client/firefox/firefox-112.0.1.ebuild
index cea794ea0bc6..ea69370898b5 100644
--- a/www-client/firefox/firefox-112.0.1.ebuild
+++ b/www-client/firefox/firefox-112.0.1.ebuild
@@ -460,9 +460,11 @@ virtwl() {
debug-print "${FUNCNAME}: $@"
"$@"
+ local r=$?
[[ -n $VIRTWL_PID ]] || die "tinywl exited unexpectedly"
exec {VIRTWL[0]}<&- {VIRTWL[1]}>&-
+ return $r
}
pkg_pretend() {
(In reply to Alex Xu (Hello71) from comment #14) > So, there's two problems here: > > 1. instrumented firefox crashes in Servo_AuthorStyles_Flush trying to write > to libxul.so .rodata. I'm still investigating this issue. > > 2. virtwl has a bug, it doesn't propagate the return code. for this issue, > please apply this patch: > > diff --git a/www-client/firefox/firefox-112.0.1.ebuild > b/www-client/firefox/firefox-112.0.1.ebuild > index cea794ea0bc6..ea69370898b5 100644 > --- a/www-client/firefox/firefox-112.0.1.ebuild > +++ b/www-client/firefox/firefox-112.0.1.ebuild > @@ -460,9 +460,11 @@ virtwl() { > > debug-print "${FUNCNAME}: $@" > "$@" > + local r=$? > > [[ -n $VIRTWL_PID ]] || die "tinywl exited unexpectedly" > exec {VIRTWL[0]}<&- {VIRTWL[1]}>&- > + return $r > } > > pkg_pretend() { Hello Alex, Thanks for investigating. I've patched the ebuild as mentioned and tried to build firefox again. However, still no luck with the same error. But i guess in order to get it built the crashes in Servo_AuthorStyles_Flush must be fixed too. Anything i can do to help? The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=6df3faf82239bb42832e1057188b5204be986415 commit 6df3faf82239bb42832e1057188b5204be986415 Author: Joonas Niilola <juippis@gentoo.org> AuthorDate: 2023-04-23 14:10:31 +0000 Commit: Joonas Niilola <juippis@gentoo.org> CommitDate: 2023-04-23 14:20:14 +0000 www-client/firefox: bug fixes for 112.0.1 - add error-state returning to virtwl(), - allow disabling telemetry via an USE flag, - include an upstream patch to fix weird bitmap font behaviour (upstream bug #1827950). Bug: https://bugs.gentoo.org/904354 Closes: https://bugs.gentoo.org/904756 Signed-off-by: Joonas Niilola <juippis@gentoo.org> www-client/firefox/Manifest | 2 +- www-client/firefox/firefox-112.0.1.ebuild | 17 ++++++++++++++--- 2 files changed, 15 insertions(+), 4 deletions(-) Minor update: Firefox still fails for me the same as before (last tried just a few minutes ago). In the meantime i've also build firefox with pgo disabled. This works so far. pgo however is still broken. iamnr3, I wasn't clear, but I meant for juippis to apply the patch, not you. it doesn't fix the build, it just makes it stop in src_compile instead of falling through to src_install. Unfortunately, I haven't had time to debug the crash, and don't think I'll have time in the next few days, but the summary is that it can be reproduced by: 1. emerge firefox 2. tinywl & 3. (. /tmp/portage/www-client/firefox-112.0.1/temp/environment; gdb /tmp/portage/www-client/firefox-112.0.1/work/firefox_build/instrumented/dist/bin/firefox) it segfaults in Servo_AuthorStyles_Flush trying to write to libxul.so .rodata. I couldn't identify the cause from the disassembly due to aggressive LLVM inlining, and I didn't have time/space yet to rebuild it with debug symbols (it uses a lot more RAM and disk). based on the short bug CC list, it seems like other users may not be encountering this, so I suspect it may be related to some specific configuration of rust, clang, etc. (In reply to Alex Xu (Hello71) from comment #18) > iamnr3, I wasn't clear, but I meant for juippis to apply the patch, not you. > it doesn't fix the build, it just makes it stop in src_compile instead of > falling through to src_install. > > Unfortunately, I haven't had time to debug the crash, and don't think I'll > have time in the next few days, but the summary is that it can be reproduced > by: > > 1. emerge firefox > 2. tinywl & > 3. (. /tmp/portage/www-client/firefox-112.0.1/temp/environment; gdb > /tmp/portage/www-client/firefox-112.0.1/work/firefox_build/instrumented/dist/ > bin/firefox) > > it segfaults in Servo_AuthorStyles_Flush trying to write to libxul.so > .rodata. I couldn't identify the cause from the disassembly due to > aggressive LLVM inlining, and I didn't have time/space yet to rebuild it > with debug symbols (it uses a lot more RAM and disk). based on the short bug > CC list, it seems like other users may not be encountering this, so I > suspect it may be related to some specific configuration of rust, clang, etc. Thanks Alex for the explanation. If I can help you somehow please let me know. I couldn't find the root cause, but seems like it is working for me now on firefox 114.0 with clang 16.0.5. it's also stable interactively for me now, 113.0.1 with USE=-pgo was constantly crashing. try 114 and if it works we'll call it "fixed"? I have PGO on and I keep getting it stalling out like it can't connect to some server, and the whole process just freezes up and I have to cancel it. I'll try again after doing some other updates and rebooting to see if there's any changes since it seems hit or miss but I did already reboot after this happened last night and still the same thing. UPDATE: After rebooting it seemed to freeze at the same point again but this time I just forgot about it and wandered off and eventually ~7-8 minutes later whatever PGO thing it was doing timed out/all the pipes were broken and it kept going, ultimately succeeding. Very odd behavior though. (In reply to anna from comment #22) > UPDATE: After rebooting it seemed to freeze at the same point again but this > time I just forgot about it and wandered off and eventually ~7-8 minutes > later whatever PGO thing it was doing timed out/all the pipes were broken > and it kept going, ultimately succeeding. Very odd behavior though. I do wonder, when this happens, whether pgo gets really enabled or not. IIRC the code dies after a certain timeout, but it could just be one area that's skipped and not the whole pgo process. Anyway if I had to guess, you're building with GCC? Try setting the clang use flag with pgo, it _should_ work better. I do have the clang USE flag set, for what it's worth. I've just installed the latest firefox with pgo. It does work for me now too. I'm fine with closing this, i guess regarding the odd behavior it would be better to open a new issue? The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=1fd128403bb5b3b4619b07e4964f7dfff635f41c commit 1fd128403bb5b3b4619b07e4964f7dfff635f41c Author: Joonas Niilola <juippis@gentoo.org> AuthorDate: 2023-08-01 10:03:51 +0000 Commit: Joonas Niilola <juippis@gentoo.org> CommitDate: 2023-08-01 10:04:27 +0000 www-client/firefox: add 116.0 Closes: https://bugs.gentoo.org/910689 Bug: https://bugs.gentoo.org/910290 Closes: https://bugs.gentoo.org/888055 Bug: https://bugs.gentoo.org/909724 Closes: https://bugs.gentoo.org/904354 Signed-off-by: Joonas Niilola <juippis@gentoo.org> www-client/firefox/Manifest | 101 +++ www-client/firefox/firefox-116.0.ebuild | 1466 +++++++++++++++++++++++++++++++ 2 files changed, 1567 insertions(+) |
