This is the mail archive of the libc-alpha@sources.redhat.com 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: libc error


bemis <bemis@iol.unh.edu> writes:

> Greeting Andreas,
>
>     So are you saying that we should ignore all libc check errors on alpha or just
> these ones?  We provide a build everyday so people that do not have the opportunity

All test-float,test-double,test-ldouble, test-ifloat, test-idouble and
test-ldobule failures will be ignored for glibc 2.2 unless you come up
with evidence that something is really broken.  

For glibc mainline, an analysis is needed why it fails and in that
case the .out file is part of the analysis.  Without that file we
cannot help.  Just telling us that test-float fails has no value at
all.

> to build on alphas can see their code built on alpha.  If we discontinue this
> autobuild system some people will not know if their code has broken libc on alpha.

If you do such kind of automatic testing, then tell the cronjob to
send the results to this list!  If you get new failures, report them
with enough analysis but the email from Mike is not a proper bug
report.

> Do you ignore compiler/check errors on intel systems that run Suse linux???

In SuSE - on all supported platforms - we test the result, analyze
them and take appropriate actions.  One action might be to ignore the
result, another might be to enhance/fix the testsuite.

>
>     I thought libc is supposed to be portable on different platforms?  Am I wrong?

Read the description of the tests in the glibc sources.  glibc is
portable but numeric algorithms might not be stable and we get tiny -
or big - differences.  The tests calculate them once and we test
against this precalculated value.

> Dejagnu has the capability to test different compilers on different platforms.  Those
> small differences are covered by dejagnu testsuites that "should" be written for each
> cpu, and its different compilers.

We're not using dejagnu in glibc,

Andreas

>
> matt
>
>
>
> Andreas Jaeger wrote:
>
>> Mike Cleary <mcleary@iol.unh.edu> writes:
>>
>> > Hello,
>> > I have found the following error with libc:
>> >
>> > GCONV_PATH=/mnt/u9/toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/native/libc/
>> > iconvdata LC_ALL=C
>> > /mnt/u9/toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/na
>> > tive/libc/elf/ld-linux.so.2 --library-path
>> > /mnt/u9/toolchain/buildsys/build/alph
>> > a9/basic-gcc_2.95.3/native/libc:/mnt/u9/toolchain/buildsys/build/alpha9/basic-gc
>> > c_2.95.3/native/libc/math:/mnt/u9/toolchain/buildsys/build/alpha9/basic-gcc_2.95
>> > .3/native/libc/elf:/mnt/u9/toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/nati
>> > ve/libc/dlfcn:/mnt/u9/toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/native/li
>> > bc/nss:/mnt/u9/toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/native/libc/nis:
>> > /mnt/u9/toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/native/libc/rt:/mnt/u9/
>> > toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/native/libc/resolv:/mnt/u9/tool
>> > chain/buildsys/build/alpha9/basic-gcc_2.95.3/native/libc/crypt:/mnt/u9/toolchain
>> > /buildsys/build/alpha9/basic-gcc_2.95.3/native/libc/linuxthreads
>> > /mnt/u9/toolcha
>> > in/buildsys/build/alpha9/basic-gcc_2.95.3/native/libc/math/test-float
>> >> /mnt/u9
>> > /toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/native/libc/math/test-float.ou
>> > t
>> > make[2]: ***
>> > [/mnt/u9/toolchain/buildsys/build/alpha9/basic-gcc_2.95.3/native/li
>> > bc/math/test-float.out] Error 1
>> >
>> > I was wondering if someone can help me solve this error.  I am finding
>> > this error on an alpha with the alphaev56-unknown-linux-gnu
>> > configuration, the basic-gcc_2.95.3 baseline, and the libc package.
>>
>> Ignore it, those are just small differences due to different compilers
>> and CPUs.
>>
>> Andreas
>> --
>>  Andreas Jaeger
>>   SuSE Labs aj@suse.de
>>    private aj@arthur.inka.de
>>     http://www.suse.de/~aj
>
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj


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