Andrew Savchenko

Indeed it would be the great feature to have: half of my mails are in UTF-8 (non-ascii), so while mairix is a great softrawe, it can't be very useful without...

Yes, that's why I connected it to #19. Looks like there there're no bug dependencies here on github, so connection is our best.

What's hard in using nanosleep()? Though I have no objections on switching this one to v0.3 schedule.

Looks like such approach will introduce some overhead, because we'll have to use clock_gettime() instead of time(), nanosleep instead of sleep and somewhat more complicated math. I wrote some test...

OTOH we already use struct timeval in notifyenginefunct.wait handlers.

Its a false positive, at least with printf sample you provided.

Let me make life easier for you: this happens due to kill() syscall: ``` $ comm clsync.failed clsync.working alarm capget capset clone close exit_group futex gettimeofday kill lstat64 mkdir mmap2...

Last trace snippets of failed process: unlink("/tmp/clsync-9biL88/.clsync-exclist.7176.3072039296.1432340755.553455361") = 0 unlink("/tmp/clsync-9biL88/.clsync-list.7176.3072039296.1432340755.944461669") = 0 tgkill(7176, 7177, SIGINT) = 0 close(0) = 0 futex(0xb7764000, FUTEX_WAIT, 2, NULL) = 0 futex(0xb7764000, FUTEX_WAKE, 1) =...

full traces are here: working: https://bpaste.net/show/6c4d79abcb0e failed: https://bpaste.net/show/df93112869a4

Meanwhile I'm worried this is a bug in the code, because even on "working" trace I see: kill(7145, SIGKILL) = -1 EPERM (Operation not permitted)