[FFmpeg-cvslog] r24156 - trunk/configure

[Måns Rullgård]

Reimar D?ffinger <Reimar.Doeffinger at gmx.de> writes:

> On Sat, Jul 10, 2010 at 11:53:07AM +0100, M?ns Rullg?rd wrote:
>> Reimar D?ffinger <Reimar.Doeffinger at gmx.de> writes:
>> 
>> > On Sat, Jul 10, 2010 at 10:26:41AM +0100, M?ns Rullg?rd wrote:
>> >> > Modified: trunk/configure
>> >> > ==============================================================================
>> >> > --- trunk/configure	Sat Jul 10 02:40:59 2010	(r24155)
>> >> > +++ trunk/configure	Sat Jul 10 06:08:02 2010	(r24156)
>> >> > @@ -2271,7 +2271,7 @@ case $target_os in
>> >> >          objformat="win32"
>> >> >          enable dos_paths
>> >> >          check_cflags -fno-common
>> >> > -        if ! enabled x86_64; then
>> >> > +        if ! check_cpp_condition _mingw.h "defined (__MINGW64_VERSION_MAJOR)"; then
>> >> 
>> >> Does the other x86_64 fail on mingw?  Are pointers 32-bit?
>> >
>> > I think you misunderstood the purpose: This is for compiling
>> > 32 bit Windows programs using the mingw64 headers/toolchain.
>> 
>> I understand the purpose but not the mechanism you are trying to
>> implement.
>
> First it is not my change, second no you don't.
>
>> Earlier on, there is this check:
>> 
>>         check_cc <<EOF && subarch="x86_64"
>>         int test[(int)sizeof(char*) - 7];
>> EOF
>> 
>> This is meant to detect whether we are building for x86_32 or x86_64
>> by looking at the pointer size.  When building for x86_32, surely
>> pointers are 32-bit, even if using a ming64 compiler.  The change
>> above should thus not be necessary.  Please explain why it makes a
>> difference so we can fix it properly.
>
> WTF? This has not the slightest bit to do with this change.

The change replaced a check for x86_64 with a different check.

> The change is about avoiding an incorrect
> die "ERROR: MinGW runtime version must be >= 3.15."
> When compiling a 32 bit binary using mingw-w64 headers.

When compiling a 32-bit binary, "enabled x86_64" should never be
true.  The change only makes any sense at all if x86_64 is incorrectly
enabled at some point.

-- 
M?ns Rullg?rd
mans at mansr.com