reentrant functions
Denys Vlasenko
vda.linux at googlemail.com
Tue Jun 10 07:30:04 UTC 2008
On Tuesday 10 June 2008 07:23, Robin Getz wrote:
> If you are talking about deviating from the "just recompiling" mantra - even
> to handle corner case error conditions, I need to rethink using uClibc going
> forward.
>
> The expectation is functionality as specified in IEEE Std 1003.1
> http://www.opengroup.org/onlinepubs/009695399/
I don't see the point to repeat again that in this case, in my opinion,
it does not matter.
We are simply disagreeing on this.
I have no problem with sometimes having a minority opinion.
Since majority opinion seems to be to not use malloc, it can be reverted.
> It seems like more and more functions that you are modifying are deviating
> from the standard, defined interface.
More and more? I don't remember changing anything else in "standard-deviating"
way, modulo bugs (that getpot embarrasment). Other changes were falling
into two categories:
1. Reduce memory usage ("smallints")
2. Fix three-year old bugs from uclibc bugzilla and failing uclibc testsuite.
> I always thought the project goals
> were:
> 1) standard functionality
> 2) minimum size
>
> Have things changed so it is the other order? Changing the project goals is
> something I would not support.
Perhaps I am personally trying to push "minimum size" more because of busybox.
It's an itch for me when I see 70k of memory wasted.
--
vda
More information about the uClibc
mailing list