apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jim Jagielski <...@jaguNET.com>
Subject Re: [PATCH] %lld support in apr_snprintf()
Date Thu, 16 Dec 2010 12:48:35 GMT

On Dec 15, 2010, at 5:11 PM, Dan Poirier wrote:

> On Mon. 2010-11-08 at 10:27 AM EST, Jim Jagielski <jim@jaguNET.com> wrote:
>> On Nov 7, 2010, at 7:42 PM, Jeff Trawick wrote:
>>> On Sun, Nov 7, 2010 at 6:51 PM, Chris Knight
>>> <Christopher.D.Knight@nasa.gov> wrote:
>>>> Exactly, the problem only appears on 64-bit Snow Leopard. See my patch in
Bugzilla, which I've verified. (Unsure if the below would also work, been a long time since
I diagnosed.)
>>> What I understood was that %lld is supposed to work, independent of
>>> which APR_*_FMT defines use %lld.
>>> I haven't tried the other patches on Leopard-64 to see which help there.
>> If one forces *just* 64bit, then, afaict, the patch is not needed.
>> It's only if one builds APR with both i386 and x86_64 that
>> things break...
> That's not my experience.  If I build just 64bit (CC="gcc -arch
> x86_64"), testfmt fails on APR_OFF_T_FMT.  (And httpd fails most of its
> byterange tests.)
> With Sander's patch to notice the second "l" in "%ll?", that test
> works but another one, using APR_UINT64_T_FMT, fails.  I haven't dug
> into that one yet.

Maybe we should make 64bit the default for APR (and httpd)
and have the weirdness only when 32bit is used. In other words,
we patch APR that the expectation is that 10.6 will be 64bit
and when compiled as 64bit, all tests and builds are clean.
It's only when 32bit, or, even-worse, a universal build, is
done that we get the errors.

It sure would be nice if when Apple releases 10.7 (and
the complimentary XCode) that this issue goes away...
View raw message