[PATCH] Remove eloop-threshold.h

Adhemerval Zanella Netto adhemerval.zanella@linaro.org
Wed Mar 19 17:21:17 GMT 2025



On 18/03/25 23:36, DJ Delorie wrote:
> Adhemerval Zanella <adhemerval.zanella@linaro.org> writes:
>> On both Linux and Hurd the __eloop_threshold() is always a constant
>> (40 and 32 respectively), so there is no need to always call
>> __sysconf (_SC_SYMLOOP_MAX) for Linux case (!SYMLOOP_MAX).  To avoid
>> a name clash with gnulib, rename the new file min-eloop-threshold.h.
> 
> LGTM with two caveats:
> 
> 1. We still #include <eloop-threshold.h> despite deleting it.  I assume
> this is for the case where it's part of gnulib, although we neither make
> that explicit nor list stdlib/canonicalize.c as shared with gnulib
> (gnulib's copy is significanly different than ours).

I change the name to min-eloop-threshold.h exactly to not mess with
gnulib definition (the eloop-threshold.h inclusion is just for !_LIBC
now).

The gnulib implementation is in fact lib/canonicalize-lgpl.c (it is not
obvious, I will try to add patch to make it explicit).

> 
> 2. eloop-threshold.h is also mentioned here:
> 
> ./benchtests/strcoll-inputs/filelist#en_US.UTF-8:	eloop-threshold.h 
> ./benchtests/strcoll-inputs/filelist#en_US.UTF-8:	eloop-threshold.h 
> 
> This will have to be fixed or the benchtests will fail.
> 
> I'd add "we should add a test to prove your assertion remains true in
> the future" but the repercussions of getting it wrong are negligible.
> 
> Reviewed-by: DJ Delorie <dj@redhat.com>
> 



More information about the Libc-alpha mailing list