[mythtv-users] mac build

[John Pilkington]

On 02/02/2021 13:52, Mark Kendall wrote:
> On Tue, 2 Feb 2021 at 00:08, James Linder <jam at tigger.ws> wrote:
>>
>> Looks like a good start point, say if you want more:
>>
>> * thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x16c13e008)
>>    * frame #0: 0x00000001044689a8 myth`AudioOutputGraph::GetImage(long long) const + 2008
>> myth`AudioOutputGraph::GetImage:
>> ->  0x1044689a8 <+2008>: movswl (%rax), %ecx
>>      0x1044689ab <+2011>: movl   %ecx, -0x168(%rbp)
>>      0x1044689b1 <+2017>: cmpl   $0x1, -0x154(%rbp)
>>      0x1044689b8 <+2024>: jle    0x1044689d4               ; <+2052>
>>
> If I were a betting man, I'd guess that:-
> 
> https://github.com/MythTV/mythtv/blob/fixes/31/mythtv/libs/libmyth/audio/audiooutputgraph.cpp#L287
> 
> is the problem. e.g. the timestamps are bogus (due to the problematic
> stream) and the error is not picked up and/or checked, we then convert
> that timestamp to an offset into a buffer - and don't bounds check the
> offset in:-
> 
> https://github.com/MythTV/mythtv/blob/fixes/31/mythtv/libs/libmyth/audio/audiooutputgraph.cpp#L123
> 
> However - I would expect the issue to be 100% repeatable if that were
> the case...
> 
> James - when it doesn't crash - do you sometimes see garbage in the
> audio graph or does it look 'normal'?
> 
> regards
> Mark

I had an encounter with the AOG code at Trac #12901, but gave up.  IIRC 
the audiotimecode there is not rigidly connected to the value at which 
the display routine is entered...

John P