This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc 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: Good mailing list to use to send out test results to


On 15/07/16 08:21, Florian Weimer wrote:
> On 07/15/2016 04:07 AM, Andrew Pinski wrote:
>> Hi,
>>   I am doing a build automation of glibc on aarch64-linux-gnu (but not
>> using the buildbot scripts) and was wondering what is a good mailing
>> list where I can send the test results summary to.
> 
> Before that, we need to figure out what's wrong with the build environment. :-/
> 
>> A sample email would be:
>>
>> Subject: Glibc revision 2b6dbe669fa2e488b31286150e8cb6f7c0875847
>> testresults for aarch64-linux-gnu
>> FAIL: conform/POSIX/glob.h/linknamespace
> 
> So far, we have seen this with excessive parallelization.  I stared about the makefile for a while, but could
> not identify the cause, so it's still not fixed.
> 
>> FAIL: debug/tst-backtrace5
>> FAIL: debug/tst-backtrace6
> 
> These are real failures, see the thread “tst-backtrace failures on AArch64” from 2014.
> 

yes

see
https://sourceware.org/glibc/wiki/Release/2.23#AArch64
for the known issues.

> The following need investigation:
> 
>> FAIL: malloc/tst-malloc-thread-exit
>> FAIL: malloc/tst-malloc-thread-fail
>> FAIL: math/test-double
>> FAIL: math/test-double-finite
>> FAIL: math/test-float
>> FAIL: math/test-float-finite
>> FAIL: math/test-idouble
>> FAIL: math/test-ifloat
>> FAIL: nptl/tst-cond16
>> FAIL: nptl/tst-cond17
>> FAIL: nptl/tst-cond20
>> FAIL: nptl/tst-dlsym1
>> FAIL: nptl/tst-robust8
>> FAIL: nptl/tst-robustpi8
>> FAIL: nptl/tst-rwlock16
>> FAIL: nptl/tst-sem14
>> FAIL: nptl/tst-stack4
>> FAIL: nptl/tst-thread_local1
>> FAIL: nss/tst-nss-static
>> FAIL: rt/tst-mqueue6
> 
> The math failures may or may not go away with ULP regeneration.  The *robust* failures may be spurious.  The
> others may be genuine issues. Unfortunately, there is not sufficient detail in the output to start an
> investigation.
> 

i'll update the libm ulps, that should fix math.

i think i've seen some of the nptl failures when
libstdc++ or libgcc_s was missing from the path.

> With such a larger number of failures, such build reports are less useful than they could be.
> 
> Florian
> 


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