[FFmpeg-devel] Patch: icl linker errors in cmdutils

[Matt Oliver]

On 8 February 2014 00:50, Derek Buitenhuis <derek.buitenhuis at gmail.com>wrote:

> On 2/7/2014 1:35 PM, Matt Oliver wrote:
> > Depending on what version of the compiler you are using /Qipo should be
> > enabled with /O2. But if you want to explicitly check it then just pass
> > /Qipo to the Windows Intel compiler so something like this should do it:
> > --extra-cflags=/Qipo
>
> I have tested with /O2 (which is what we set) on both archs, on both
> supported ICL
> versions, as listed at:
>
>
> http://ffmpeg.org/platform.html#Microsoft-Visual-C_002b_002b-or-Intel-C_002b_002b-Compiler-for-Windows
>
> Using any other version, with your own broken CFLAGS is not a supported
> use case.
>
> Either that, or you are not giving a proper config line for me to test.
>
> - Derek
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>

Well I dont use the extra cflags option as I was having the automated Qipo
without specifying it. I assumed that on your version you might have to
manually specify ipo (FYO its ipo not Qipo on linux) to reproduce the error
but apparently not.
I can try and make a test case for you if you really need to have it but
given the patch doesnt brake anything then even if this error doesnt affect
the build tools you have the worst case scenario is that this patch removes
some unneeded if(0) and if(1)s and slightly increases compilation times as
a result. I figure with a patch as simple as this that has no obvious down
sides the exact nature of the reproducible error shouldn't be of as much
concern. Clearly there is an issue as the code stands (after alll why would
I make a patch for an issue that doesnt exist) so whats the harm in fixing
it?