svn commit: r299456 - in head: include lib/libc/stdio

Conrad Meyer cem at FreeBSD.org
Thu May 12 22:50:36 UTC 2016


On Thu, May 12, 2016 at 12:49 PM, Antoine Brodin <antoine at freebsd.org> wrote:
> On Wed, May 11, 2016 at 4:38 PM, Conrad E. Meyer <cem at freebsd.org> wrote:
>> Author: cem
>> Date: Wed May 11 14:38:27 2016
>> New Revision: 299456
>> URL: https://svnweb.freebsd.org/changeset/base/299456
>>
>> Log:
>>   libc: Add fopencookie(3) wrapper around funopen(3)
>>
>>   Reviewed by:  jhb, oshogbo
>>   Sponsored by: EMC / Isilon Storage Division
>>   Differential Revision:        https://reviews.freebsd.org/D6282
>
> Please revert this and request a ports exp-run as this breaks lots of
> important ports.

The number of failed ports is actually pretty small (less than 11).
I've identified a couple different kinds of issue and implemented
fixes for them.

1. libc doesn't actually export the fopencookie() function, because it
was missing from Symbols.map (base issue).  This is rectified in
299572 and I've personally test-built php56 successfully with that
change.

2. Some ports assume that off64_t's presence in stdio.h implies it is
more widely available (port issue).  Or that it implies __off64_t
exists.  Examples: devel/rudiments, devel/zziplib.  To that end, I've
made off64_t and __off64_t as widely available as off_t and __off_t in
r299571.  I've test-built both of these and both succeed after these
changes.

3. Finally, Clang (?)sometimes rejects that particular style of
function typedef (base issue).  Additionally, sometimes size_t isn't
defined.  For example, lang/pypy.  r299574 fixes that issue.  I've
personally test-built pypy (well, it's still compiling, but it's past
that initial platform detection failure now).

Are there any other important ports you think this change impacted?
Several other failures are not related to this change (Lynx, Xterm,
ncurses, cdialog).  So, other than assuming php70 and php55 will be
fixed by the same change as php56, I don't think there are any
failures unaccounted for that I haven't test-built successfully at
this point.

Do you still want a revert, or any further test builds?  Or is that fine?

Best,
Conrad


More information about the svn-src-head mailing list