Monday Patch Queue Review update (2021-11-29)

Monday Patch Queue Review update (2021-11-29)

[Carlos O'Donell via Libc-alpha]

Most recent meeting status is always here:
https://sourceware.org/glibc/wiki/PatchworkReviewMeetings#Update

Meeting: 2021-06-28 @ 0900h EST (UTC-5)

IRC: #glibc on OFTC.

 * 354 patches in NEW/NOBODY
 * [v3] inet: Return EAI_MEMORY when nrl_domainname() fails to allocate memory
  * Doesn't apply. Does it need a patch before this?
 * linux: Use /proc/stat fallback for __get_nprocs_conf (BZ #28624)
  * Florian to provide additional feedback about simplification.
 * x86-64: Add vector math functions to libmvec
  * Need v2 posted rebased on top of other bench changes.
 * [v2] Add GLIBC_ABI_VERSION_GEN2 for ABIVERSION
  * Some confusion over the behaviour of these changes.
  * Florian to post some questions.
 * linux: Some syscall refactors
  * Stafford and Adhemerval reviewing.
 * 47844: [v3] elf: Pre-generate Makefile fragment for DSO sorting tests [BZ #28550]
  * Needs more comments.
 * AArch64: Optimize memcmp
  * Szabolcs to reivew.
 * Stoped at 47810.
 * Siddhesh: Archival rules for patchwork instances.
  * November 15th email thread to libc-alpha.
  * What is the tool for this?
  * Carlos will respond with feedback for archival rules.
  * https://sourceware.org/pipermail/libc-alpha/2021-November/133125.html
 * Paul Zimmerman: llvm-libc developers
  * Want to implement correctly rounded routines.
 * 46928: [v2] Remove upper limit on tunable MALLOC_MMAP_THRESHOLD
  * Patrick responded to DJ. Waiting for feedback from DJ.
 * Florian hopes to post a new version of the rseq integration.
  * 48162: [v2] elf: Add _dl_find_object function
   * Needs further review. Carlos wondered if Szabolcs (Arm) could provide feedback.

-- 
Cheers,
Carlos.

Re: Monday Patch Queue Review update (2021-11-29)

[Florian Weimer via Libc-alpha]

* Carlos O'Donell via Libc-alpha:

>  * Florian hopes to post a new version of the rseq integration.
>   * 48162: [v2] elf: Add _dl_find_object function
>    * Needs further review. Carlos wondered if Szabolcs (Arm) could provide feedback.

Thanks for posting the minutes.  Actually I forgot to mention another
patch of mine:

  [PATCH v2] elf: Install a symbolic link to ld.so as /usr/bin/ld.so
  <https://sourceware.org/pipermail/libc-alpha/2021-November/133104.html>

It's a simple change, but I would like to receive some community
feedback if this direction is okay.

Florian

Re: Monday Patch Queue Review update (2021-11-29)

[Adhemerval Zanella via Libc-alpha]

On 30/11/2021 11:37, Carlos O'Donell via Libc-alpha wrote:
> Most recent meeting status is always here:
> https://sourceware.org/glibc/wiki/PatchworkReviewMeetings#Update
> 
> Meeting: 2021-06-28 @ 0900h EST (UTC-5)
> 
> IRC: #glibc on OFTC.
> 
>  * 354 patches in NEW/NOBODY
>  * [v3] inet: Return EAI_MEMORY when nrl_domainname() fails to allocate memory

This is because I sent a version against a patchset updating only the last patch,
which confuses patchwork.  I will just resend the whole set.

>   * Doesn't apply. Does it need a patch before this?
>  * linux: Use /proc/stat fallback for __get_nprocs_conf (BZ #28624)
>   * Florian to provide additional feedback about simplification.

I think Florian remarks are in fact a different issue, where we assume sysfs
support (the issue was in fact a patch of mine removed too much).

>  * x86-64: Add vector math functions to libmvec
>   * Need v2 posted rebased on top of other bench changes.
>  * [v2] Add GLIBC_ABI_VERSION_GEN2 for ABIVERSION
>   * Some confusion over the behaviour of these changes.
>   * Florian to post some questions.
>  * linux: Some syscall refactors
>   * Stafford and Adhemerval reviewing.
>  * 47844: [v3] elf: Pre-generate Makefile fragment for DSO sorting tests [BZ #28550]
>   * Needs more comments.
>  * AArch64: Optimize memcmp
>   * Szabolcs to reivew.
>  * Stoped at 47810.
>  * Siddhesh: Archival rules for patchwork instances.
>   * November 15th email thread to libc-alpha.
>   * What is the tool for this?
>   * Carlos will respond with feedback for archival rules.
>   * https://sourceware.org/pipermail/libc-alpha/2021-November/133125.html
>  * Paul Zimmerman: llvm-libc developers
>   * Want to implement correctly rounded routines.
>  * 46928: [v2] Remove upper limit on tunable MALLOC_MMAP_THRESHOLD
>   * Patrick responded to DJ. Waiting for feedback from DJ.
>  * Florian hopes to post a new version of the rseq integration.
>   * 48162: [v2] elf: Add _dl_find_object function
>    * Needs further review. Carlos wondered if Szabolcs (Arm) could provide feedback.
> 

Re: Monday Patch Queue Review update (2021-11-29)

[Florian Weimer via Libc-alpha]

* Adhemerval Zanella via Libc-alpha:

>>   * Doesn't apply. Does it need a patch before this?
>>  * linux: Use /proc/stat fallback for __get_nprocs_conf (BZ #28624)
>>   * Florian to provide additional feedback about simplification.
>
> I think Florian remarks are in fact a different issue, where we assume sysfs
> support (the issue was in fact a patch of mine removed too much).

I think we should have changed

  return 1;

to

  return 2;

in __get_nprocs, and not to:

  return get_nproc_stat (buffer, buffer_size);

It would have fixed the reported issue because we would keep
conservatively reporting SMP support.

Thanks,
Florian

Re: Monday Patch Queue Review update (2021-11-29)

[Adhemerval Zanella via Libc-alpha]

On 30/11/2021 13:19, Florian Weimer wrote:
> * Adhemerval Zanella via Libc-alpha:
> 
>>>   * Doesn't apply. Does it need a patch before this?
>>>  * linux: Use /proc/stat fallback for __get_nprocs_conf (BZ #28624)
>>>   * Florian to provide additional feedback about simplification.
>>
>> I think Florian remarks are in fact a different issue, where we assume sysfs
>> support (the issue was in fact a patch of mine removed too much).
> 
> I think we should have changed
> 
>   return 1;
> 
> to
> 
>   return 2;
> 
> in __get_nprocs, and not to:
> 
>   return get_nproc_stat (buffer, buffer_size);
> 
> It would have fixed the reported issue because we would keep
> conservatively reporting SMP support.

But it is doing exactly this right now, since get_nproc_stat will
return 2 if /proc can not be opened.