This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: QMTest Update


In article <66640000.1022542881@warlock.codesourcery.com>,
Mark Mitchell<mark@codesourcery.com> writes:

> Thank you for trying out QMTest last week.

Mark, I was happy to do so.  Thank you for being able to decode my
sketchy bug report.

> You reported some problems with multithreaded operation which puzzled
> me at first -- but then I found a system that behaved as you described.
> There were some unportable constructs in the GCC test classes. [...]

I can confirm that you have fixed the problem for my environment.
There is a minor (unrelated?) nit remaining.  I manually killed one
test run under -j2 after it ran for an hour.  It "properly" failed
under -j1.  This produced the obvious change in the final report:

5c5,6
<       14 (  0%) tests unexpected FAIL
---
>        1 (  0%) tests unexpected ERROR
>       13 (  0%) tests unexpected FAIL
51,52c52,53
<   gpp.old-deja.jason.template26                 : FAIL    , expected PASS    
<     Executable received fatal signal 139.
---
>   gpp.old-deja.jason.template26                 : ERROR   , expected PASS    
>     An exception occurred.

It might also be nice if the output recorded in gpp.qmr was stable
across -j settings and the presence or lack of --seed.  Alternatively,
there might already be XML tools that could process said output into a
comparable sequence.  Of course, the only reason I even want that
property is so I may continue to use standard UNIX text tools where
XML-based tools are perhaps available.

Regards,
Loren


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]