Re: accommodating OS/x, bleeding edge gcc and warnings

[Eric Blake]

On 10/28/2013 02:35 PM, Eric Blake wrote:

> Given the Comments in vfprintf-posix:
>> Comment:
>> This module should not be used as a dependency from a test module,
>> otherwise when this module occurs as a tests-related module, it will
>> have side effects on the compilation of the 'vasnprintf' module, if
>> that module occurs among the main modules in lib/.
> 
> and the fact that LGPL projects such as libvirt can end up using error
> from the testsuite, I'm inclined to revert this patch.  Can you provide
> more evidence of why you thought it was necessary, so that we can fix
> the real root cause of the problem you encountered without dragging in
> vfprintf into the testsuite and corrupting main lib/ clients of
> vasnprintf?

Is the real problem that gnulib's error.h uses merely:

__attribute__((__format__))

while gnulib's stdio.h replacement is careful to use either
__attribute__((__format__ __gnu_printf__)) or the weaker
__attribute__((__format__ __printf__)) depending on what other gnulib
*printf[-posix] modules are in effect?  That is, were you hitting a
compilation warning on OS/x due to a mismatch between error's
declaration being too weak for your version of gcc's detection?

If so, then the real fix should be to lib/error.h, not to modules/error.
Is it okay if I revert your earlier patch in the meantime?

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature