On 02/21/2011 04:24 PM, Jeffrey Altman wrote:
On 2/21/2011 11:56 AM, Marc Dionne wrote:
On Mon, Feb 21, 2011 at 11:16, Andrew Deason<[email protected]> wrote:
On Mon, 21 Feb 2011 08:41:47 -0500
Marc Dionne<[email protected]> wrote:
Maybe we could detect in the build script whether the current commit
touches any configure files (src/cf/*, *.m4, *.ac) and include it only
in that case? Using "git diff" creatively can probably give us that
information.
git diff-tree --numstat --no-commit-id HEAD | cut -f3
It still seems only likely to be useful if the build fails, though,
isn't it?
Sure, but that's probably true of most of the buildbot logs - I would
think that not many people scan any of the logs for builds that
succeed. Limiting it to changes that touch files that affect the
configure step should already make it apply to a small percentage of
commits. Surely we can afford the extra overhead for those cases.
But if it can also be limited to failed builds I'd be OK with that.
Marc
I scan the log for the addition of warnings in sections of code on
platforms which I do not have access to.
Well, I reverted the config.log change fr now until I can implement a
better way.
I understand that config.log is useful if the ./configure step fails. Is
there a case where config.log is useful if ./configure succeeds?
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel
