1204822 - No such file libgtk-3.so.0 in Hazard build

Status: RESOLVED FIXED

(Testing :: General, defect)

Description

[John Dai[:jdai]]

According to bug 1146713 comment 150, B2G Desktop Linux x64 Hazard Analysis Build fail and got  /bin/bash: /builds/slave/b2g_b2g-in_l64-b2g-haz_dep-000/build/gaia/b2g_sdk/43.0a1-2015-08-17-15-02-06/b2g/xpcshell: No such file or directory

Comment 1

[John Dai[:jdai]]

Attachment: Add LD_LIBRARY_PATH in build script.

Moving set LD_LIBRARY_PATH value code to buildbase.py. B2G Desktop Linux x64 Hazard Analysis Build success.

Try link:
https://treeherder.allizom.org/#/jobs?repo=try&revision=9fa12bbac4eb

Hi Dustin, 
May I have your review? Thanks.

Comment 2

[Dustin J. Mitchell [:dustin] (he/him)]

Comment on attachment 8661152 [details] [diff] [review]
Add LD_LIBRARY_PATH in build script.

This looks good, but because it's in buildbase I'm worried it will affect other builds, so I've flagged Jordan for a second review.

Comment 3

[John Dai[:jdai]]

Comment on attachment 8661152 [details] [diff] [review]
Add LD_LIBRARY_PATH in build script.

Cancel the review first. I want to check Bug 1204800 will pass build also.

Comment 4

[John Dai[:jdai]]

Attachment: Add LD_LIBRARY_PATH in build script. v2

Verified on all builds, they are all green.

Try link:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=6fcbd2f2489a

Comment 5

[Jordan Lund (:jlund)]

Comment on attachment 8662198 [details] [diff] [review]
Add LD_LIBRARY_PATH in build script. v2

Review of attachment 8662198 [details] [diff] [review]:
-----------------------------------------------------------------

should be harmless. only b2g device and hazard use B2GBuildBaseScript which I guess is the hazard motivation for putting the libgtk-3 logic there.

Comment 6

[John Dai[:jdai]]

Attachment: Add LD_LIBRARY_PATH in build script. r=dustin, jlund

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/a59090148268

Comment 8

[Nigel Babu [:nigelb]]

This has caused an interesting issue (see bug 1207009 for actual failure). 

cmd = ["ln", "-s", gtk3_path, gtk3_symlink_path]

This line causes a fatal error that causes the build to be reported as a failure when the symblink exists. The build will succeed once on all slaves and from the second time, it will throw an error like so

17:58:43     INFO - Running main action method: build_symbols
17:58:43     INFO - Running command: ['ln', '-s', '/builds/slave/b2g_m-in_emu_dep-0000000000000/build/gtk3', '/builds/slave/b2g_m-in_emu_dep-0000000000000/build/gecko/gtk3']
17:58:43     INFO - Copy/paste: ln -s /builds/slave/b2g_m-in_emu_dep-0000000000000/build/gtk3 /builds/slave/b2g_m-in_emu_dep-0000000000000/build/gecko/gtk3
17:58:43     INFO -  ln: creating symbolic link `/builds/slave/b2g_m-in_emu_dep-0000000000000/build/gecko/gtk3/gtk3': File exists
17:58:43    ERROR - Return code: 1
17:58:43    ERROR - failed to create symlink

We should probably force the symlinking or fail gracefully if the link already exists. For now, I'm backing this out so that B2G ICS builds turn green again.

Comment 9

[Pulsebot]

Backout:
https://hg.mozilla.org/integration/mozilla-inbound/rev/cadeefe3e927

Comment 10

[John Dai[:jdai]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=100043a3121f

Comment 11

[John Dai[:jdai]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=aa35c45796c9

Comment 12

[John Dai[:jdai]]

Attachment: Add LD_LIBRARY_PATH in build script. r=dustin, jlund

Use the symlinking if the link not exists.

Try link:
https://treeherder.allizom.org/#/jobs?repo=try&revision=60bd5048e3ab

Comment 13

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/f4233421a009

Comment 14

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

Backed out for possibly causing asan/mulet test failures like https://treeherder.mozilla.org/logviewer.html#?job_id=14582981&repo=mozilla-inbound

https://hg.mozilla.org/integration/mozilla-inbound/rev/ff14d1e4beea

Comment 15

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

Tests were still broken on the backout, so this likely wasn't at fault. I'll reland it later today once I sort out what did break it.

Comment 16

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/74ec968ef9c5

Comment 17

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/74ec968ef9c5