Diego Biurrun <[email protected]> writes:

> On Sat, Sep 08, 2012 at 03:37:18PM +0300, Martin Storsjö wrote:
>> On Sat, 8 Sep 2012, Måns Rullgård wrote:
>> >Hendrik Leppkes <[email protected]> writes:
>> >>On Sat, Sep 8, 2012 at 1:40 PM, Martin Storsjö <[email protected]> wrote:
>> >>>On Sat, 8 Sep 2012, Måns Rullgård wrote:
>> >>> Hendrik Leppkes <[email protected]> writes:
>> >>>> On Sat, Sep 8, 2012 at 12:09 PM, Ronald S. Bultje <[email protected]
>> >>>>>>wrote:
>> >>>>>> This sounds like the wrong fix tbh. The first chunk looks OK but
>> >>>>>>wouldn't other systems, eg old GCC versions, be similarly affected?
>> >>>>>>Shouldn't the true fix live elsewhere, ie fix the log2() check?
>> >>>>>>
>> >>>>>The log2 check is fine in theory. The linker somewhere finds a log2
>> >>>>>function to link in, but without a function declaration in the header.
>> >>>>
>> >>>>Find out where that log2 is coming from and what it really is.
>> >>>
>> >>>It seems that the log2 is present in the static libc that MSVC normally
>> >>>uses (but not in the dynamically loaded one), and the headers don't 
>> >>>contain
>> >>>any declaration of it.
>> >>>
>> >>And only in the 64-bit version. How odd.
>> >>
>> >>Still, something goes wrong when it trys to use it, causing fate to fail.
>> >
>> >If you have an implicit declaration for it returning int, it can't
>> >possibly work.
>> 
>> It seems to work fine if a manual declaration of the function
>> (double log2(double);) is added.
>
> So how are we going to resolve this?

I see two simple solutions:

1. Add a declaration for log2() in a suitable place.  This is slightly
   evil in that it uses an undocumented function.  Then again, that
   appears to be expected behaviour on Windows.

2. Add log2_deps="!msvcrt" somewhere in configure.  This would wrongly
   disable the function if they were ever to add it for real.

Pick your poison.

-- 
Måns Rullgård
[email protected]
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to