Closed Bug 1900590 Opened 9 months ago Closed 8 months ago

Undefined symbol aom_codec_control first referenced in file ../../../image/decoders/Unified_cpp_image_decoders0.o

Categories

(Core :: Audio/Video: Web Codecs, defect)

Product:
Core
Component:
Audio/Video: Web Codecs
Type:
defect

Tracking

()

Status:
RESOLVED FIXED
Milestone:
129 Branch
Tracking Flags:
Tracking Status
firefox-esr115 --- unaffected
firefox127 --- unaffected
firefox128 --- fixed
firefox129 --- fixed

People

(Reporter: petr.sumbera, Assigned: chunmin)

References

(Regression)

Details

(Keywords: regression)

Attachments

(1 file)

Bug 1900590 - Suppress undefined aom_codec_control symbol on Solaris SPARC
48 bytes, text/x-phabricator-request
RyanVM
: approval-mozilla-beta+
Details | Review

Steps to reproduce:

Build on Solaris SPARC failed with:

30:38.49 security/manager/ssl/ipcclientcerts/dynamic-library/libipcclientcerts.so
35:14.49 Undefined                      first referenced
35:14.49  symbol                            in file
35:14.49 aom_codec_control                   ../../../image/decoders/Unified_cpp_image_decoders0.o  (symbol scope specifies local binding)
35:14.49 ld: fatal: symbol referencing errors
35:17.93 collect2: error: ld returned 1 exit status
35:17.94 gmake[4]: *** [/builds/psumbera/mozilla-central-build/config/rules.mk:541: ../../../dist/bin/libxul.so] Error 1

Seems to be regression from Bug 1876438.

Keywords: regression
Regressed by: 1876438

The Bugbug bot thinks this bug should belong to the 'Firefox Build System::General' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Component: Untriaged → General
Product: Firefox → Firefox Build System

:chunmin, since you are the author of the regressor, bug 1876438, could you take a look? Also, could you set the severity field?

For more information, please visit BugBot documentation.

Flags: needinfo?(cchang)

Hi Petr, Thanks for the report. Does running ./mach clobber before ./mach build help? aom_codec_control has been exposed from gkcodecs for a while. I am not sure why it causes problem in image/decoders now. IIUC, the nsAVIFDecoder is the only one file that touch aom library. Could you try adding USE_LIBS += [ 'gkcodecs' ] to image/decoders/moz.build to see if it solves the issue?

Flags: needinfo?(cchang) → needinfo?(petr.sumbera)
Component: General → Audio/Video: Web Codecs
Product: Firefox Build System → Core

I do build from clean workspace. No need for clobber.

In the meantime my hg bissect on I386 Solaris confirmed:

The first bad revision is:
changeset:   741427:f089bb5743b1
user:        Chun-Min Chang <chun.m.chang@gmail.com>
date:        Mon Jun 03 11:45:58 2024 -0700
description:
	Bug 1876438 - Enable SVC configuration in FFmpegVideoEncoder for libaom r=media-playback-reviewers,padenot

I have tried:

diff -r 8c97559c3a94 -r a26ba0f04542 image/decoders/moz.build
--- a/image/decoders/moz.build  Thu Oct 05 00:56:26 2023 -0700
+++ b/image/decoders/moz.build  Wed Jun 05 00:25:47 2024 -0700
@@ -58,5 +58,7 @@

 CXXFLAGS += ["-Werror=switch"]

+USE_LIBS += [ 'gkcodecs' ]
+
 # Add libFuzzer configuration directives
 include("/tools/fuzzing/libfuzzer-config.mozbuild")

But it doesn't make difference.

Flags: needinfo?(petr.sumbera)

I see that aom_codec_control is defined in aom_codec.o

nm obj-x86_64-pc-solaris2.11/media/libaom/aom_codec.o | grep aom_codec_control$
[21]    |                 558|                 485|FUNC |GLOB |0    |1      |aom_codec_control

but I don't see anywehere libgkcodecs.so...

obj-x86_64-pc-solaris2.11/toolkit/library/build/libxul_so.list is missing libaom items...

When I revert:

Bug 1876438 - Enable SVC configuration in FFmpegVideoEncoder for libaom r=media-playback-reviewers,padenot

I can still build latest mozilla-central sources.

Questions remains. How shall I make libxul to link libaom objects...

Mike, I wonder whether you have an idea how obj-x86_64-pc-solaris2.11/media/libaom/*.o are supposed to get into libxul_so.list so that they are linked into libxul?

Flags: needinfo?(mh+mozilla)

media/libaom/*.o are not supposed to get into libxul_so.list. They're supposed to get into libgkcodecs_so.list.

Flags: needinfo?(mh+mozilla)

Thank you. Now I have at least limited test case. Though not really sure where is the problem (libgkcodecs.so is included).

$ /usr/gcc/13/lib/gcc/x86_64-pc-solaris2.11/13.2.0/collect2 -G -dy -z text -M /usr/gcc/13/lib/gcc/x86_64-pc-solaris2.11/13.2.0/../../../libgcc-unwind.map -Y P,/lib/amd64:/usr/lib/amd64 -Qy -o ../../../dist/bin/libxul.so -z gnu-version-script-compat /usr/lib/amd64/crti.o /usr/gcc/13/lib/gcc/x86_64-pc-solaris2.11/13.2.0/crtbeginS.o -L/builds/psumbera/epoll -L/builds/psumbera/mozilla-central-build/obj-x86_64-pc-solaris2.11/dist/bin -L/usr/gcc/13/lib/gcc/x86_64-pc-solaris2.11/13.2.0 -L/usr/gcc/13/lib/gcc/x86_64-pc-solaris2.11/13.2.0/../../../amd64 -L/lib/amd64 -L/usr/lib/amd64 -L/usr/gcc/13/lib/gcc/x86_64-pc-solaris2.11/13.2.0/../../..  -lepoll -h libxul.so @/builds/psumbera/mozilla-central-build/obj-x86_64-pc-solaris2.11/toolkit/library/build/libxul_so.list -z text -z now ../../../js/src/build/libjs_static.a ../../../build/pure_virtual/libpure_virtual.a ../../../x86_64-pc-solaris/debug/libgkrust.a ../../../dist/bin/libnspr4.so ../../../dist/bin/libplc4.so ../../../dist/bin/libplds4.so ../../../dist/bin/libgkcodecs.so ../../../dist/bin/liblgpllibs.so ../../../dist/bin/libnss3.so ../../../dist/bin/libnssutil3.so ../../../dist/bin/libsmime3.so ../../../dist/bin/libmozsqlite3.so ../../../dist/bin/libssl3.so ../../../dist/bin/libmozgtk.so --version-script libxul.so.symbols -lsocket -ldl -lpthread -lz -lposix4 -lnsl -lfreetype -lfontconfig -lgtk-3 -lgdk-3 -lpangocairo-1.0 -lpango-1.0 -lharfbuzz -latk-1.0 -lcairo-gobject -lcairo -lgdk_pixbuf-2.0 -lgio-2.0 -lgobject-2.0 -lglib-2.0 -levent -ldbus-1 -lxcb-shm -lX11-xcb -lX11 -lxcb -lXext -lXrandr -lstdc++ -lm -lssp_nonshared -lssp -lgcc_s -lc -lgcc_s /usr/gcc/13/lib/gcc/x86_64-pc-solaris2.11/13.2.0/crtendS.o /usr/lib/amd64/crtn.o
Undefined                       first referenced
 symbol                             in file
aom_codec_control                   ../../../image/decoders/Unified_cpp_image_decoders0.o  (symbol scope specifies local binding)
ld: fatal: symbol referencing errors
collect2: error: ld returned 1 exit status
$ nm ../../../dist/bin/libgkcodecs.so | grep aom_codec_control$
[82503] |            11992814|                 485|FUNC |GLOB |0    |19     |aom_codec_control

I was suggested to use some Solaris linker debugging (https://www.mail-archive.com/tools-linking@mail.opensolaris.org/msg00573.html).

aom_codec_control is a normal global symbol looking for a definition, when it encounters an instance from Unified_cpp_ffmpeg_libav530.oUnified_cpp_ffmpeg_libav530.o that has hidden visibility specified, and this causes ld to give the resulting merged symbol hidden visibility:

debug: symbol[5987]=aom_codec_control  (global); resolving twoundefs
debug:        old  0x0000000000000000 0x0000000000000000  NOTY GLOB  D    0 UNDEF          ../../../image/decoders/Unified_cpp_image_decoders0.o
debug:        new  0x0000000000000000 0x0000000000000000  NOTY GLOB  H    0 UNDEF          ../../../dom/media/platforms/ffmpeg/libav53/Unified_cpp_ffmpeg_libav530.o
debug:   resolved  0x0000000000000000 0x0000000000000000  NOTY GLOB  H    0 UNDEF          REF_REL_NEED  (Unified_cpp_image_decoders0.o)

The question is why that hidden visibility is being requested, and what effect the author is hoping to get from doing so..

Sounds like some header missing in config/system-headers.mozbuild. It's surprising this doesn't cause problems on Linux, though.

Maybe we should add aom_encoder.h in system_headers? aom_codec_control is defined in "aom/aom_codec.h", and "aom/aom_codec.h" is used via in <aom/aom_encoder.h> in libaomenc.c

(In reply to C.M.Chang[:chunmin] from comment #13)

Maybe we should add aom_encoder.h in system_headers? aom_codec_control is defined in "aom/aom_codec.h", and "aom/aom_codec.h" is used via in <aom/aom_encoder.h> in libaomenc.c

This didn't make a difference:

--- a/config/system-headers.mozbuild    Thu Oct 05 00:56:26 2023 -0700
+++ b/config/system-headers.mozbuild    Thu Oct 05 00:56:26 2023 -0700
@@ -1030,6 +1030,7 @@
     "aom/aom_image.h",
     "aom/aomdx.h",
     "aom/aom_decoder.h",
+    "aom/aom_encoder.h",
 ]

 if CONFIG["MOZ_X11"]:

Note that there is no system version on libaom library and no system headers like /usr/include/aom.

Severity: -- → S3

Uff, I wish I understand it. Following change resolves the issue. Thank you for your hints!

diff -r b6762d206c10 -r 8e7ecb951dd3 config/system-headers.mozbuild
--- a/config/system-headers.mozbuild    Thu Oct 05 00:56:26 2023 -0700
+++ b/config/system-headers.mozbuild    Thu Oct 05 00:56:26 2023 -0700
@@ -1029,6 +1029,7 @@
     "vpx_mem/vpx_mem.h",
     "aom/aom_image.h",
     "aom/aomdx.h",
+    "aom/aomcx.h",
     "aom/aom_decoder.h",
 ]

Will you please push it to mozilla-cetrall? Ideally also to FF 128 branch.

Flags: needinfo?(cchang)

This patch fixes the undefined aom_codec_control symbol on Solaris
SPARC by adding "aom/aomcx.h" to config/system-headers.mozbuild.
Thanks to Petr Sumbera (petr.sumbera@oracle.com) for identifying and
solving the issue [1].

[1] https://bugzilla.mozilla.org/show_bug.cgi?id=1900590#c16

(In reply to Petr Sumbera from comment #16)

Will you please push it to mozilla-cetrall? Ideally also to FF 128 branch.

Hi Petr, thanks for working out a fix! I've pushed a try to see if it breaks anything, and will land it soon!
https://treeherder.mozilla.org/jobs?repo=try&revision=10c91e85d067cdeb7e43bd403be57cc49bd87477

Flags: needinfo?(cchang)
Assignee: nobody → cchang
Status: UNCONFIRMED → ASSIGNED
Ever confirmed: true

Set release status flags based on info from the regressing bug 1876438

status-firefox127: --- → unaffected
status-firefox128: --- → affected
status-firefox129: --- → affected
status-firefox-esr115: --- → unaffected

(In reply to C.M.Chang[:chunmin] from comment #18)

Hi Petr, thanks for working out a fix! I've pushed a try to see if it breaks anything, and will land it soon!
https://treeherder.mozilla.org/jobs?repo=try&revision=10c91e85d067cdeb7e43bd403be57cc49bd87477

Just for record. The problem is on both SPARC and i386 Solaris. Probably some Solaris linker specific feature.

Pushed by cchang@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/c338d4f02f98 Suppress undefined aom_codec_control symbol on Solaris SPARC r=glandium

https://hg.mozilla.org/mozilla-central/rev/c338d4f02f98

Status: ASSIGNED → RESOLVED
Closed: 8 months ago
status-firefox129: affected → fixed
Resolution: --- → FIXED
Target Milestone: --- → 129 Branch

The patch landed in nightly and beta is affected.
:chunmin, is this bug important enough to require an uplift?

  • If yes, please nominate the patch for beta approval.
  • If no, please set status-firefox128 to wontfix.

For more information, please visit BugBot documentation.

Flags: needinfo?(cchang)

Comment on attachment 9408320 [details]
Bug 1900590 - Suppress undefined aom_codec_control symbol on Solaris SPARC

Beta/Release Uplift Approval Request

  • User impact if declined: Solaris SPARC and i386 Solaris cannot build Firefox
  • Is this code covered by automated tests?: Yes
  • Has the fix been verified in Nightly?: Yes
  • Needs manual test from QE?: No
  • If yes, steps to reproduce:
  • List of other uplifts needed: None
  • Risk to taking this patch: Low
  • Why is the change risky/not risky? (and alternatives if risky): This patch is just a fix for solving the build issue on Solaris SPARC and i386 Solaris
  • String changes made/needed:
  • Is Android affected?: No
Flags: needinfo?(cchang)
Attachment #9408320 - Flags: approval-mozilla-beta?

Comment on attachment 9408320 [details]
Bug 1900590 - Suppress undefined aom_codec_control symbol on Solaris SPARC

Approved for 128.0b6.

Attachment #9408320 - Flags: approval-mozilla-beta? → approval-mozilla-beta+
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: