Closed GoogleCodeExporter closed 8 years ago
See also build bot failures
buildbot failure in Libyuv on Android Release
The Buildbot has detected a new failure on builder Android Release while
building libyuv.
Full details are available at:
http://build.chromium.org/p/client.libyuv/builders/Android%20Release/builds/716
Buildbot URL: http://build.chromium.org/p/client.libyuv/
Buildslave for this Build: vm322-m3
Build Reason: scheduler
Build Source Stamp: [branch master] 2560df9513e66080524a5b8dcea92b8ec657a8eb
Blamelist: fbarchard@google.com
BUILD FAILED: failed steps failed gclient runhooks
sincerely,
-The Buildbot
Original comment by fbarch...@google.com
on 5 Jan 2016 at 11:17
I had a look at the bot and it was failing due to some file system permission
issues that was possible to resolve by wiping a
/mnt/data/b/build/slave/android/build/src/chromium/src/third_party/android_tools
on disk. So both commit bots in
https://build.chromium.org/p/client.libyuv/waterfall are now green.
However, I'm also able to reproduce the error in my local checkout, so I wonder
what the difference can be. The bots also gets auto-updated depot tools, and it
doesn't look related to depot tools to me, rather something with GYP.
Original comment by kjellan...@google.com
on 10 Jan 2016 at 8:02
OK I've figured it out: at
https://code.google.com/p/chromium/codesearch#chromium/src/build/android/gyp/jav
a_cpp_enum.py&l=233 the script uses the absolute path to lookup the name of the
script (in a very odd and probably bad way). If 'build' is not a part of the
path, it simply fails.
The works in Chromium and WebRTC since the GYP scripts (build/gyp_chromium and
webrtc/build/gyp_webrtc) are located in such paths. For libyuv, the gyp_libyuv
is put straight into the root of the checkout, thus not containing 'build'.
The reason it passes on the libyuv buildbots is just pure luck, since their
checkouts are located in a dir like: /b/build/slave/linux64/build/libyuv.
I'll see if I can make a bugfix for the java_cpp_enum.py script so this error
goes away.
Another option would be to try to find a way to not invoke that GYP stuff, but
it's hard since libyuv_tests.gyp includes testing/android/native_test.gyp,
which is where this comes from.
Workaround: Make a directory 'build' and move your libyuv workspace into it :)
Original comment by kjellan...@google.com
on 10 Jan 2016 at 8:27
CL up: https://codereview.chromium.org/1571243003/
Original comment by kjellan...@google.com
on 11 Jan 2016 at 4:30
build/ folder work around works, but roll forward causes a build error?
../../third_party/gflags/src/gflags.cc:281:53: error: unused typedef
'true_false_equal' [-Werror,-Wunused-local-typedef]
COMPILE_ASSERT(sizeof(kTrue) == sizeof(kFalse), true_false_equal);
^
1 error generated.
[8/231] CXX obj/chromium/src/testing/gtest/src/gtest.gtest.o
Original comment by fbarch...@google.com
on 13 Jan 2016 at 2:38
The default warnings in Chromium for third party code has changed.
You need to update gflags GYP+GN with the changes made in
https://codereview.webrtc.org/1553033002
Original comment by kjellan...@google.com
on 13 Jan 2016 at 8:12
The following revision refers to this bug:
https://chromium.googlesource.com/libyuv/libyuv.git/+/f7c6b1a1967c89513d8d6156dc1cf42ca5a16d2c
commit f7c6b1a1967c89513d8d6156dc1cf42ca5a16d2c
Author: Frank Barchard <fbarchard@google.com>
Date: Thu Jan 14 19:09:28 2016
pull gflags.gyp changes from webrtc
gflags has a build warning in the chromium version.
../../third_party/gflags/src/gflags.cc:281:53: error: unused typedef
'true_false_equal' [-Werror,-Wunused-local-typedef]
COMPILE_ASSERT(sizeof(kTrue) == sizeof(kFalse), true_false_equal);
^
1 error generated.
[8/231] CXX obj/chromium/src/testing/gtest/src/gtest.gtest.o
Its been reported and fixed upstream
https://github.com/gflags/gflags/issues/125
But not pulled into chromium. Last update to flags was in 2013.
This change disables the warning.
R=harryjin@google.com, kjellander@chromium.org, jkellander@chromium.org
BUG=libyuv:536, webrtc:5397, webrtc:5398, webrtc:5399
Review URL: https://codereview.chromium.org/1581123002 .
[modify]
http://crrev.com/f7c6b1a1967c89513d8d6156dc1cf42ca5a16d2c/third_party/gflags/BUI
LD.gn
[modify]
http://crrev.com/f7c6b1a1967c89513d8d6156dc1cf42ca5a16d2c/third_party/gflags/gfl
ags.gyp
Original comment by bugdroid1@chromium.org
on 14 Jan 2016 at 7:09
The following revision refers to this bug:
https://chromium.googlesource.com/libyuv/libyuv.git/+/d1c6eb0887e83fb03e00aa9c35be59a26e4735a0
commit d1c6eb0887e83fb03e00aa9c35be59a26e4735a0
Author: Frank Barchard <fbarchard@google.com>
Date: Thu Jan 14 19:39:13 2016
Roll chromium to fix java_cpp_enum.py
Remove assumption on absolute paths in java_cpp_enum.py.
In java_cpp_enum.py there were assumptions on 'build' being
a component of the absolute path of the script invoking the
script. This leads to problems when other projects are using
the Chromium build toolchain (in this case libyuv), where
the script invoking GYP is located in the root of the project
dir (https://chromium.googlesource.com/libyuv/libyuv/+/master/gyp_libyuv).
TBR=harryjin@google.com, jkellander@chromium.org
BUG=libyuv:536
Review URL: https://codereview.chromium.org/1581773003 .
[modify] http://crrev.com/d1c6eb0887e83fb03e00aa9c35be59a26e4735a0/DEPS
Original comment by bugdroid1@chromium.org
on 14 Jan 2016 at 7:39
Original comment by fbarch...@google.com
on 16 Jan 2016 at 2:13
Original issue reported on code.google.com by
fbarch...@google.com
on 22 Dec 2015 at 1:02