This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Race condition in localedata tests?
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>, Konstantin Serebryany <kcc at google dot com>
- Cc: GNU C Library <libc-alpha at sourceware dot org>
- Date: Tue, 18 Nov 2014 14:46:53 -0500
- Subject: Re: Race condition in localedata tests?
- Authentication-results: sourceware.org; auth=none
- References: <CAMe9rOp1ytEWzT8NzBj9-UtG1Ev0378BUj=PsGN2AwNSVa2KMA at mail dot gmail dot com> <546B73C4 dot 2020505 at redhat dot com> <CAMe9rOrPZMSvG4tBgLLATKrGh3Kj-S-1ost20TVOGJX+_097Fg at mail dot gmail dot com> <CAMe9rOpDD3847qjaRSuBvuFRgfyQ9Ks0yDxg=JgaTWbj+GbfYQ at mail dot gmail dot com>
On 11/18/2014 11:41 AM, H.J. Lu wrote:
>>> Do the tests have explicit dependencies on the locales they use?
>>> Perhaps it just happens that de_DE.UTF-8 hits the filesystem by
>>> the time the test runs.
>>>
>>
>> localedata/Makefile has
>>
>> # Dependency for the locale files. We actually make it depend only on
>> # one of the files.
>> $(addprefix $(objpfx),$(CTYPE_FILES)): %: \
>> gen-locale.sh $(common-objpfx)locale/localedef Makefile \
>> $(addprefix charmaps/,$(CHARMAPS)) $(addprefix locales/,$(LOCALE_SRCS))
>> @$(SHELL) gen-locale.sh $(common-objpfx) \
>> '$(built-program-cmd-before-env)' '$(run-program-env)' \
>> '$(built-program-cmd-after-env)' $@; \
>> $(evaluate-test)
>>
>> $(addsuffix .out,$(addprefix $(objpfx),$(tests))): %: \
>> $(addprefix $(objpfx),$(CTYPE_FILES))
>>
>> $(objpfx)sort-test.out: sort-test.sh $(objpfx)collate-test $(objpfx)xfrm-test \
>> $(test-input-data) $(addprefix $(objpfx),$(CTYPE_FILES))
>> $(SHELL) $< $(common-objpfx) '$(test-program-prefix-before-env)' \
>> '$(run-program-env)' '$(test-program-prefix-after-env)' \
>> $(test-input) \
>> > $@; \
>> $(evaluate-test)
>>
>> However, there is no $(CTYPE_FILES) dependency for those tests
>> above, which fail at random under heavy load with parallel build.
>>
>>
>
> I opened:
>
> https://sourceware.org/bugzilla/show_bug.cgi?id=17617
H.J.,
Thanks. I expected this might be the case. We'd seen other issues with
high-level parallelism failing.
Kostya,
Given that you're the only other person I know who runs the glibc testsuite
at high levels of parallelism, was there anything else you saw in your testing
that fell over? I know you mentioned locales once. I know some of the tests
don't parallelize well, but that's not a failure per se.
Cheers,
Carlos.