On Sun, 8 Jul 2012, Ronald S. Bultje wrote:
Hi,
On Sun, Jul 8, 2012 at 12:48 PM, Måns Rullgård <m...@mansr.com> wrote:
"Ronald S. Bultje" <rsbul...@gmail.com> writes:
Hi,
On Sat, Jul 7, 2012 at 3:57 PM, Derek Buitenhuis
<derek.buitenh...@gmail.com> wrote:
On 07/07/2012 6:09 PM, Diego Biurrun wrote:
And this is required - why?
Libm / math is not A Thing in MSVC and isn't required.
It complains about -lm, but ignores it.
I think this is mostly to shut up a warning.
Correct. First, link.exe does not use -l<libname> to link in
libraries, it uses lib<name>.lib (similar to how gcc allows you to
specify the full filename of a lib<name>.a static library). Second,
there is no math lib necessary at all.
This is the wrong fix for that problem. The correct fix is to translate
-lfoo to the supported syntax. This is required anyway, and once done
the original test will work exactly as intended.
Wrong thread. Please comment only on the -lm piece which isn't
necessary in this thread. I'd like -lm to not exist on systems not
needing it (e.g. OS2, Haiku, MSVC).
On OSes that don't need it but that interpret -lm correctly, it won't be
added. As he pointed out in another mail, testing for one math function
without -lm isn't enough since that particular function might be inlined -
testing without -lm would require testing every single math function we
use.
The general "-lname" -> "libname.lib" conversion discussion should go
in the other thread.
No, if we fix that, configure will notice that adding -lm (which renders
adding 'm.lib' to the command line options) fails, and thus won't add it -
just as intended.
// Martin
_______________________________________________
libav-devel mailing list
libav-devel@libav.org
https://lists.libav.org/mailman/listinfo/libav-devel