On 04/09/2012 01:37 PM, Steven Jenkins wrote:
On Mon, Apr 9, 2012 at 1:34 PM, Simon Wilkinson
<[email protected]> wrote:
On 9 Apr 2012, at 15:03, Jason Edgecombe wrote:
hi everyone,
I'm thinking of adding a "make check" step to the non-windows build slaves.
Is this a good idea? Bad idea? Comments?
Theoretically, it is a great idea. There are a couple of potential problems,
though.
*) Ideally, make check would do a complete test of all of the functionality in OpenAFS.
Obviously we are a very long way away from that, but each test that we add will slow down
the "make check" process. I have worked on large projects where the test suite
takes much, much longer to execute than the build process, so this is a consideration for
our slower builders.
*) I'm not sure that all of our tests fail cleanly. There may be some tests
that fail by hanging, or some failure modes that leave daemons running. Dealing
with these will add to the administration overhead of the build farm.
I think I'd prefer, at the moment, that we had a single builder doing a make
check, rather than having every builder execute them.
Cheers,
Another option is to have multiple 'check' targets, eg., 'make
check:build' suitable for buildbots, 'make check:comprehensive' for a
more comprehensive, but slower, test suite.
Steven
I've added a "make check" step to the freebsd82 and debian-i386
builders. It shows as the "test" step.
I can add it to any other builders or I can add it to only the scheduled
builds.
Let me know if there are any issues or suggestions.
BTW, the "make check" fails on BSD and Linux.
Jason
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel
