[llvm] r272021 - [lit] Improve logging with file redirection.

[Hans Wennborg via llvm-commits]

On Thu, Jun 9, 2016 at 9:24 PM, Daniel Dunbar <daniel at zuster.org> wrote:
> I don't see any obvious problem where we would leak an FD, but maybe this
> exposed underlying deadlock conditions (which the code already calls out are
> possible in some cases).

We might not even have been really leaking them, but the change might
have pushed how many we have open at the same time over the limit
somehow.

> I fixed this logic to only capture the output when there are actual
> failures, which was my original intent anyway, and my guess is that this
> will help.
>
> Can you retest the failing scenario after r272365 and let me know if you
> still see it?

I tried with r272365, and now everything works again :-) Thanks!

 - Hans

> On Thu, Jun 9, 2016 at 6:21 PM, Hans Wennborg <hans at chromium.org> wrote:
>>
>> On Thu, Jun 9, 2016 at 1:36 PM, Hans Wennborg <hans at chromium.org> wrote:
>> > On Tue, Jun 7, 2016 at 9:13 AM, Daniel Dunbar via llvm-commits
>> > <llvm-commits at lists.llvm.org> wrote:
>> >> Author: ddunbar
>> >> Date: Tue Jun  7 11:13:40 2016
>> >> New Revision: 272021
>> >>
>> >> URL: http://llvm.org/viewvc/llvm-project?rev=272021&view=rev
>> >> Log:
>> >> [lit] Improve logging with file redirection.
>> >>
>> >>  - This will cause lit to automatically include the first 1K of data in
>> >>    redirected output files when a command fails (previously if the
>> >> command
>> >>    failed, but the main point of the test was, say, a `FileCheck` later
>> >> on, then
>> >>    the log wasn't helpful in showing why the command failed).
>> >
>> > This makes the test suite hang for me on 32-bit Windows builds. It
>> > runs fine at r272019, but hangs at this revision (and also on
>> > ToT/r272292).
>> >
>> > I'm not sure how to debug this. It seems after a while, python.exe
>> > just sits there waiting for something, with 1, 2, or 0 FileCheck
>> > processes opened.
>>
>> Maybe it's a problem of running out of file descriptors. I'm not sure
>> why this is happening on my machine but not on buildbots, though. If I
>> run lit with -j16, or set NUMBER_OF_PROCESSORS in the environment, the
>> tests pass, so I'm not blocked on this anymore.
>
>