Bug 83226 – build-webkit | filter-build-webkit returns success even when build-webkit fails

• Status: RESOLVED
• Resolution: WONTFIX
• Component: Tools / Tests

[Eric Seidel (no email)]

build-webkit | filter-build-webkit returns success even when build-webkit fails

Fixing this would make my life easier. :)  I would like to be able to use:

build-webkit | filter-build-webkit && run-webkit-tests

[Eric Seidel (no email)]

Hmm... Looks like bug 70080 attempted this before.

[Eric Seidel (no email)]

Another solution would be to just integrate filter-build-webkit into build-webkit :)

[Dean Jackson]

(In reply to comment #2)
> Another solution would be to just integrate filter-build-webkit into build-webkit :)

Yeah. Pass --filter-build or something.

[Eric Seidel (no email)]

mdelany doesn't seem to have a bugzilla account or I would have CC'd him. :)
http://trac.webkit.org/log/trunk/Tools/Scripts/filter-build-webkit

[Blaze Burg]

There is nothing that we can change in filter-build-webkit (a perl script) to get the exit status from an earlier command in the pipeline.

There are two ways to get this behavior without changing any scripts:

1. In bash, use cmd1 |& cmd2 (shortcut for cmd1 2>&1 | cmd2). Then the BUILD FAILED detection will work.
2. In bash, run 'set -o pipefail'. This uses the leftmost non-zero return code from a pipeline rather than the rightmost program.

In the future, build-webkit should incorporate filter-build-webkit. If filter-build-webkit were running build commands, then it could capture and propagate the exit codes from its child processes.