r239402 - Remove rm invocations where the file is immediately rewritten later.

Aaron Ballman aaron at aaronballman.com
Tue Jun 9 09:26:53 PDT 2015

On Tue, Jun 9, 2015 at 12:23 PM, Benjamin Kramer <benny.kra at gmail.com> wrote:
> On Tue, Jun 9, 2015 at 6:02 PM, Greg Bedwell <gregbedwell at gmail.com> wrote:
>>> Remove rm invocations where the file is immediately rewritten later.
>>> This may or may not help making this test less flaky on windows. There's
>>> a race condition in lit somewhere.
>> I've always assumed that this is just Windows holding onto the file for too
>> long, possibly exacerbated by the presence of AV software, rather than an
>> issue with lit itself although I'm just wildly speculating based on previous
>> experience with this sort of thing.  I'm still hitting this failure on my
>> machine occasionally today after this commit although I'll have to use it
>> for a few days more before deciding whether it reduces the frequency of the
>> failure or not.
>> If it really is just a Windows thing rather than a lit thing I'm not sure
>> what we'd be able to do beyond something like having lit intercept calls to
>> rm and perform some sort of retry method for permission errors on Windows.
>> I'm pretty sure that will work, but it's extremely nasty, and if the problem
>> is more heinous than just Windows holding onto files it's not a real fix
>> either.
>> Thanks for having looked at this either way though!  It's a relief to know
>> it's not just me that keeps running into this one.
> Right, it's still failing intermittently :(
> Indexer/AV seems like a more likely explanation and one that's going
> to be hard to fix. Just working around 'rm' would fix this specific
> tests, but there are others (archive-update.test) that also fail on
> windows from time to time and that one seems unrelated to rm.
> I rarely run tests on windows myself (takes forever) but I want the
> buildbots to stop yelling at me ...

Finding a solution to this would be excellent. It fails for me on
Windows fairly regularly, but all attempts to debug it seem to make
the issue disappear for me (also suggesting a timing issue). It's
really aggravating. :-/


More information about the cfe-commits mailing list