From: Jonathan W. <jw...@ju...> - 2011-09-27 23:45:34
|
> > ======= Backtrace: ========= > > /lib/x86_64-linux-gnu/libc.so.6(+0x78a96)[0x7fa8b80dba96] > > /lib/x86_64-linux-gnu/libc.so.6(cfree+0x6c)[0x7fa8b80dfd7c] > > /usr/lib/libffado.so.2(_ZN11DebugModuleD0Ev+0xa2)[0x7fa8ba8b7bc2] > > /usr/lib/libffado.so.2(_ZN18DebugModuleManagerD1Ev+0x9e)[0x7fa8ba8b745e] > > This is #329: > > http://subversion.ffado.org/ticket/329 Yep. > Tracing the underlying problem when using ordinary binutils is also > another option. I have recently upgraded my development system which gives me binutils 2.20.51.0.8.20100412. Do we know which version seemed to be the flag day for this bug? > So I don't think we're seeing a regression with r1995 ... I agree, it's not technically a regression, although users may see it as such. With reference to r1995, the only change in this revision was the addition of mutex initialisers in the constructors which really matter. Until this was done those mutexes were being used uninitialised which is why some people were seeing crashes during the isohandler closedown. There's no way the correct initialisation of this mutex could be directly causing the debug module to crash. I can well imagine, however, that differing memory layouts as a result of r1995 could well make the debugmodule crash easier to trigger for some people. But that would just be due to the nature of the bug underlying ticket 329 rather than being directly caused by r1995. Regards jonathan |