Re: size_max

[Simon Josefsson]

Eric Blake <address@hidden> writes:

> Simon Josefsson <simon <at> josefsson.org> writes:
>
>> So how about this patch?
>> 
>> Of course, stdint.h has to define SIZE_MAX on these systems to make sure
>> it will still work.  But I suppose we'll notice if it doesn't already.
>
> stdint.in.h already takes care of SIZE_MAX.

Yes but I recall some reports about SIZE_MAX problem anyway, but it may
because of config.h or size_max.m4 issues.  My patch appears to be the
right thing, so unless anyone finds a problem with it, I'd like to move
ahead with that and sort out any problems that are reported later on.

> Fix these nits, then I think it's ready to apply...

I want to hear from Bruno first though.

>
>> 
>> +2009-10-06  size_max        The header file "size_max.h" has been removed,
>> +                            use <stdint.h> from the stdint.h module instead.
>
> s/stdint.h module/stdint module/

Fixed.

>>  Files:
>> -m4/size_max.m4
>>  lib/size_max.h
>
> Oops - you meant to stop distributing lib/size_max.h but keep m4/size_max.m4 
> since gl_SIZE_MAX is still a valid configure macro.

Heh, definitely.  Fixed.

>>  Depends-on:
>
> I debated whether we need to add stdint here.  But I'm guessing not, so that 
> the xsize module will continue to work without dragging in the stdint 
> replacement header.  So the NEWS is sufficient to tell the majority of 
> clients 
> to use the correct module in the first place.

I think so too.  Also, any application that includes stdint.h and
assumes it provides SIZE_MAX should pull in the stdint module directly.
I don't think it is the size_max's modules job.

> Then, are you also willing to prepare the followup patch to kill the
> self- definition of SIZE_MAX throughout the various modules willing to
> depend on the stdint module?

I'll work on it when we've heard from Bruno -- I recall that this was
discussed earlier, and I have paged out most of that context so there
could be reasons to not do this.

/Simon