[mythtv] Compiler issue on github runner for Fixes/32 macOS 11/12

[Scott Theisen]

On 11/23/22 18:37, John Hoyt wrote:
>
>
> On Wed, Nov 23, 2022 at 6:28 PM John Hoyt <john.hoyt at gmail.com> wrote:
>
>     Hi Scott, thanks for taking a look.  I've worked up a patch file
>     to resolve, but it sounds like getting configure working correctly
>     is a better solution (happy to post the patch file).
>
>         Without the outputs from configure, particularly the log file
>         config.ep, I can't say for certain if those lines are the problem.
>
>
>     Configure output: https://pastebin.com/gG8E1Fbg
>     config.ep: https://pastebin.com/b6Tk5M4Q
>
>

from config.ep line 5736:
```
dyld[67989]: Library not loaded: '@PREFIX@/lib/libhdhomerun.dylib'
Referenced from: 
'/private/var/folders/4c/__l40lp140324rnbz1m2jkgh0000gn/T/ffconf.xBPEc1aO/test'
Reason: tried: '/usr/local/lib/libhdhomerun.dylib' (no such file), 
'/usr/lib/libhdhomerun.dylib' (no such file)
```

That looks suspicious to me.  Would that prevent the test from executing?

The resultant config.h files from our configure and FFmpeg's would be 
useful to compare.

> BTW - I can now reproduce on a local machine (wife's macbook) running 
> Monterey.  Strangely, all of my Ventura machines do not exhibit the 
> problem.
>
> Here's the patch I worked up that successfully compiles:
>
>     diff --git a/mythtv/external/FFmpeg/libavutil/libm.h
>     b/mythtv/external/FFmpeg/libavutil/libm.h
>     index a819962391..71753deff5 100644
>     --- a/mythtv/external/FFmpeg/libavutil/libm.h
>     +++ b/mythtv/external/FFmpeg/libavutil/libm.h
>     @@ -427,21 +427,21 @@ static inline double rint(double x)
>      #endif /* HAVE_RINT */
>      #if !HAVE_LRINT
>     -static av_always_inline av_const long int lrint(double x)
>     +av_always_inline av_const long int lrint(double x)
>      {
>          return rint(x);
>      }
>      #endif /* HAVE_LRINT */
>      #if !HAVE_LRINTF
>     -static av_always_inline av_const long int lrintf(float x)
>     +av_always_inline av_const long int lrintf(float x)
>      {
>          return (int)(rint(x));
>      }
>      #endif /* HAVE_LRINTF */
>      #if !HAVE_ROUND
>     -static av_always_inline av_const double round(double x)
>     +av_always_inline av_const double round(double x)
>      {
>          return (x > 0) ? floor(x + 0.5) : ceil(x - 0.5);
>      }
>     diff --git a/mythtv/libs/libmyth/audio/audioconvert.cpp
>     b/mythtv/libs/libmyth/audio/audioconvert.cpp
>     index 98a2e25c8c..1b7903e599 100644
>     --- a/mythtv/libs/libmyth/audio/audioconvert.cpp
>     +++ b/mythtv/libs/libmyth/audio/audioconvert.cpp
>     @@ -71,7 +71,7 @@ static inline bool sse_check()
>      #endif //ARCH_x86
>      #if !HAVE_LRINTF
>     -static av_always_inline av_const long int lrintf(float x)
>     +av_always_inline av_const long int lrintf(float x)
>      {
>          return (int)(rint(x));
>      }
>
>

I would prefer to fix configure, rather than hack around the error.

Were the HAVE_LRINTF changes necessary?  I don't think they should be.

I think a better solution is to remove the referenced configure lines 
and see if that works.  My reasoning is the following:

 1. `git grep -nE "math.h" -- :^*/FFmpeg/** :^platform/**` We use
    <cmath> not <math.h>, except in external/libexiv2.
 2. We never use lrint() `git grep -nE lrint -- :^*/FFmpeg/**` only
    lrintf().
      * Although, we do use round(), `git grep -nE "[^A-Za-z]round\(" --
        :^*/FFmpeg/** :^*/html/**`.
 3. We are using C++17, these functions are from C99/C++11.

Regards,

Scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mythtv.org/pipermail/mythtv-dev/attachments/20221123/c5711e9e/attachment.htm>