Re: glibc 2.22 --- Stil frozen. Machine maintainers start testing.

["Carlos O'Donell" <carlos at redhat dot com>]

On 07/28/2015 06:47 AM, Szabolcs Nagy wrote:
> On 22/07/15 16:09, Szabolcs Nagy wrote:
>> On 20/07/15 21:15, Carlos O'Donell wrote:
>>> This is a reminder that master is still frozen.
>>>
>>> Only bug fixes and documentation patches.
>>>
>>> Machine maintainers should *start* testing to make sure you
>>> don't run into any unforseen problems.
>>>
>>> I'd like to have the last bug fixes in by the end of the week.
>>
>> aarch64 status:
> 
> i cannot edit
> https://sourceware.org/glibc/wiki/Release/2.22

Please follow:
https://sourceware.org/glibc/wiki/EditorGroup

If you tell me your wiki login name then I can add you, and you
can add your test results to the wiki.

> here is an update on a new toolchain:
> 
> gcc version 6.0.0 20150724 (experimental), aarch64-unknown-linux-gnu
> GNU ld (GNU Binutils) 2.25.51.20150727
> Linux 3.18.5 #1 SMP PREEMPT aarch64
> 
> XPASS: conform/ISO11/complex.h/conform
> XPASS: conform/ISO11/stdalign.h/conform
> XPASS: conform/ISO11/stdnoreturn.h/conform
> FAIL: debug/tst-backtrace5
> FAIL: debug/tst-backtrace6
> FAIL: nptl/tst-stack4
> Summary of test results:
>       3 FAIL
>    2252 PASS
>      88 XFAIL
>       3 XPASS
> 
>> FAIL: debug/tst-backtrace5
>> FAIL: debug/tst-backtrace6
>> there is a fix as part of the cancel changes:
>> https://sourceware.org/ml/libc-alpha/2015-06/msg00950.html
>> i can fix this independently, but i think it is ok
>> to delay this until next release and do it with the
>> cancellation fixes.
> 
> still applies.
> 
>> other pending issues:
>>
>> BZ 17918 dynamic-linker race: pthread_create can access
>> GL(dl_tls_max_dtv_idx) and GL(dl_tls_generation) without
>> holding the rtld lock.
>> https://sourceware.org/bugzilla/show_bug.cgi?id=17918
>> i don't know about any fix for this.
> 
> this is the cause of nptl/tst-stack4 failure, it affects all archs,
> it happens about once out of 100 runs here on aarch64.

Given it's an existing bug we'll try to tackle this in the more thorough
P&C cleanup for that code in 2.23.
 
> Inconsistency detected by ld.so: dl-tls.c: 493:
> _dl_allocate_tls_init: Assertion `listp->slotinfo[cnt].gen <=
> _rtld_local._dl_tls_generation' failed!
> 
>> BZ 18707 aarch64 --enable-profile is broken because asm
>> for the mcount call clobbers registers it shouldn't.
>> https://sourceware.org/bugzilla/show_bug.cgi?id=18707
> 
> i have a fix for this, not published yet.

OK.

c.