Issue 1736 in webm: Build fails with LTO enabled (GCC 11)

[sergi… via monorail]

Status: Untriaged
Owner: ----
Labels: Type-Bug Pri-2
Components: libvpx

New issue 1736 by sergi...@canonical.com: Build fails with LTO enabled (GCC 11)
https://bugs.chromium.org/p/webm/issues/detail?id=1736

Ubuntu has performed a full rebuild of the archive using GCC 11, and libvpx is FTBFS'ing:

https://launchpadlibrarian.net/552670245/buildlog_ubuntu-impish-amd64.libvpx_1.9.0-1_BUILDING.txt.gz

...
g++ -Wl,-Bsymbolic-functions -flto=auto -Wl,-z,relro -Wl,-z,now -m64 -o test_libvpx ivfenc.c.o md5_utils.c.o test/active_map_refresh_test.cc.o test/active_map_test.cc.o test/alt_ref_aq_segment_test.cc.o test/altref_test.cc.o test/aq_segment_test.cc.o test/bench.cc.o test/borders_test.cc.o test/byte_alignment_test.cc.o test/config_test.cc.o test/cpu_speed_test.cc.o test/cq_test.cc.o test/decode_api_test.cc.o test/decode_corrupted.cc.o test/decode_svc_test.cc.o test/decode_test_driver.cc.o test/encode_api_test.cc.o test/encode_test_driver.cc.o test/error_resilience_test.cc.o test/external_frame_buffer_test.cc.o test/frame_size_tests.cc.o test/invalid_file_test.cc.o test/keyframe_test.cc.o test/level_test.cc.o test/realtime_test.cc.o test/resize_test.cc.o test/svc_datarate_test.cc.o test/svc_end_to_end_test.cc.o test/svc_test.cc.o test/test_libvpx.cc.o test/test_vector_test.cc.o test/test_vectors.cc.o test/timestamp_test.cc.o test/user_priv_test.cc.o test/vp8_datarate_test.cc.o test/vp9_datarate_test.cc.o test/vp9_end_to_end_test.cc.o test/vp9_ethread_test.cc.o test/vp9_lossless_test.cc.o test/vp9_motion_vector_test.cc.o test/vp9_skip_loopfilter_test.cc.o test/y4m_test.cc.o third_party/libwebm/mkvparser/mkvparser.cc.o third_party/libwebm/mkvparser/mkvreader.cc.o webmdec.cc.o y4menc.c.o y4minput.c.o -L. -lvpx -lgtest -lpthread -lm -lpthread
ln -sf libvpx.so.6.3.0 vpx-vp8-vp9-x86_64-linux-v1.9.0/lib/libvpx.so.6
/usr/bin/ld: /tmp/ccsyaUhJ.ltrans0.ltrans.o:(.debug_info+0x1586f): undefined reference to `gtest_all.cc.5c9bdf8f'
/usr/bin/ld: /tmp/ccsyaUhJ.ltrans0.ltrans.o:(.debug_info+0x158a5): undefined reference to `gtest_all.cc.5c9bdf8f'
/usr/bin/ld: /tmp/ccsyaUhJ.ltrans0.ltrans.o:(.debug_info+0x158b2): undefined reference to `gtest_all.cc.5c9bdf8f'
/usr/bin/ld: /tmp/ccsyaUhJ.ltrans0.ltrans.o:(.debug_info+0x158b7): undefined reference to `gtest_all.cc.5c9bdf8f'
/usr/bin/ld: /tmp/ccsyaUhJ.ltrans0.ltrans.o:(.debug_info+0x158dd): undefined reference to `gtest_all.cc.5c9bdf8f'
/usr/bin/ld: /tmp/ccsyaUhJ.ltrans0.ltrans.o:(.debug_info+0x15903): more undefined references to `gtest_all.cc.5c9bdf8f' follow
collect2: error: ld returned 1 exit status
make[3]: *** [/<<PKGBUILDDIR>>/libs.mk:623: test_libvpx] Error 1
make[2]: *** [Makefile:41: dist] Error 2

The problem goes away when LTO is disabled, which is what we did for now. Please let us know if you need more to diagnose this issue.

--
You received this message because:
1. The project was configured to send all issue notifications to this address

You may adjust your notification preferences at:
https://bugs.chromium.org/hosting/settings

[sergi… via monorail]

Comment #1 on issue 1736 by sergi...@canonical.com: Build fails with LTO enabled (GCC 11)
https://bugs.chromium.org/p/webm/issues/detail?id=1736#c1

Ah, I've just tested with the current git HEAD and noticed that the problem doesn't seem to manifest there. Please disregard this report, then. Thanks, and sorry about the noise.

[sergi… via monorail]

Comment #2 on issue 1736 by sergi...@canonical.com: Build fails with LTO enabled (GCC 11)
https://bugs.chromium.org/p/webm/issues/detail?id=1736#c2

Sigh... I think I forgot to use a compilation flags and for that reason the bug didn't manifest. I can actually reproduce it using git HEAD by compiling the project using this:

LDFLAGS='-Wl,-Bsymbolic-functions -flto=auto -Wl,-z,relro' CFLAGS='-g -O2 -ffile-prefix-map=/root/libvpx/libvpx/builddir=. -flto=auto -ffat-lto-objects -fstack-protector-strong -Wformat -Werror=format-security' CXXFLAGS='-g -O2 -ffile-prefix-map=/root/libvpx/libvpx/builddir=. -flto=auto -ffat-lto-objects -fstack-protector-strong -Wformat -Werror=format-security' CPPFLAGS='-Wdate-time -D_FORTIFY_SOURCE=2' ../configure --prefix=/usr --enable-pic --enable-shared --disable-install-bins --disable-install-srcs --size-limit=16384x16384 --enable-postproc --enable-multi-res-encoding --enable-temporal-denoising --enable-vp9-temporal-denoising --enable-vp9-postproc --target=x86_64-linux-gcc && make -j8 verbose=yes dist

So it seems this bug *is* in fact valid against git HEAD.

[jz… via monorail]

Comment #3 on issue 1736 by jz...@google.com: Build fails with LTO enabled (GCC 11)
https://bugs.chromium.org/p/webm/issues/detail?id=1736#c3

Thanks for the report. Does the issue only show with the test code? Can you try adding --disable-unit-tests to your configure command line?

[sergi… via monorail]

Comment #4 on issue 1736 by sergi...@canonical.com: Build fails with LTO enabled (GCC 11)
https://bugs.chromium.org/p/webm/issues/detail?id=1736#c4

The issue doesn't manifest when --disable-unit-tests is used, indeed.

[jz… via monorail]

Comment #5 on issue 1736 by jz...@google.com: Build fails with LTO enabled (GCC 11)
https://bugs.chromium.org/p/webm/issues/detail?id=1736#c5

That's good to know. If it doesn't happen with the latest upstream version of googletest [1] we should just need to update the snapshot.

[1] https://github.com/google/googletest