Subject: Re: CVS commit: pkgsrc/security/AiSSLtelnet
To: Rhialto <firstname.lastname@example.org>
From: Krister Walfridsson <email@example.com>
Date: 07/19/2005 14:45:24
On Tue, 19 Jul 2005, Rhialto wrote:
> On Tue 19 Jul 2005 at 11:42:41 +0200, Dieter Baron wrote:
>> In article <20050719041450.8024B2DA27@cvs.netbsd.org> Krister wrote:
>> : The sources assume <malloc.h> exists. Create a fake one for platforms
>> : that don't have it.
>> Shouldn't we centralize support for this rather than duplicating it
>> in N Makefiles? Maybe a NEEDS_MALLOC_H or USE_MALLOC_H which triggers
>> the creation in mk/bsd.*.mk?
"Maybe". The reason why I haven't done that is that I agree with
Rhialto's main point below, and didn't want yet another option that
we want to remove soon.
> <malloc.h> is not a standard header file. Never has been. Any program
> requiring it is plain broken. Don't "fix" such a program by adding a
> header file, fix it by patching it to use <stdlib.h> instead (and feel
> free to send a cross email to the author).
I mostly agree, and I send patches to the authors where appropriate
(many of the malloc.h using packages are unmaintained...)
But creating lots of patches makes the package harder to update etc.
so my general strategy has been to try to make it easy to maintain:
I have added patches if it's just one/two files, or the package
never will be updated. For the others, I mail the author, and
add the fake header (which easily can be removed in the next
upgrade of the package).