After some segabrt/segfaults the download library’s time-out code has been cleaned up. It sometimes tried to destruct locked mutexes.

Minor tweaks and such have gone on but more recently I did a bit of research on forking – though my services use it I was still unclear what exactly it allows:

  • Forking creates new processes
  • Each process has the thread limit
  • Fork twice, 300 per process, total 600 threads
  • In this case, not much point as mutex locks cause everything to wait so it would just wait in another process – no performance gain
  • The powercache and paged queue need work – cant escape that.

Leave a Reply