Abramo Bagnara wrote: >Tim Goetze wrote: >> >> >>but sometimes i also get inexplicable corruption within ordinary >> >>dynamically allocated memory of the process before it exits. >> >> >> >>i have spent considerable time on verifying that these are indeed >> >>caused by changing the period_size, and not by my own code. >> > >> >i would be too certain of this conclusion. i spent many a long night >> >and day and night tracking down places in the audioengine code that >> >had overrun the mmapped areas. i was only ever to track things down by >> >generating huge log files that detailed every byte i wrote, and then >> >use perl and awk to digest them. >> > >> >i'm not saying its not a problem in ALSA, only that its unlikely at >> >this point. what have you done to ensure that its not your code? do >> >you completely avoid writing data to the mapped areas? >> > >> >> yes, exactly that. i just did >> n * (open, configure, m * (poll, mmap_begin, mmap_commit), close) >> and nothing more. > >You're missing avail_update: it's *mandatory*. > >> i ran this twice, one time with the usual configure, and one time >> with f/c hardcoded to 64 in the hw setup so my code in all places believes >> it's going 2048, 1024, 512, ... -- not that it matters because i don't >> process any data. the one that's doing the 'real configure' crashes.
yes, i did avail_update. sorry for not quoting all the 12111 lines of code involved. and before i receive another message like this: i also did snd_pcm_link, _unlink, _start and _drop. tim _______________________________________________ Alsa-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/alsa-devel
