[Bug libc/30558] SIGEV_THREAD is badly implemented

["stsp at users dot sourceforge.net" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=30558

--- Comment #4 from Stas Sergeev <stsp at users dot sourceforge.net> ---
Yes, I used your example to check with
glibc, and it appears it indeed spawns
any amount of threads concurrently if
the handler sleeps.
But in this case I think that impl is
not even posix-conforming because in
that scheme you can't use timer_getoverrun()
consistently. The man page of timer_getoverrun()
explicitly says that the timer doesn't
queue more than 1 signal, so that
timer_getoverrun() can show the overrun
count between signal queuing and dequeuing.
To me that still looks crazy because
after signal dequeuing another overrun
could occur. But at least that scheme
can be made working, because linux AFAICS
"cures" that by adding such lost overruns
upon the next signal delivery.

But if you spawn multiple threads
concurrently, how are you going to use
timer_getoverrun() at all?

-- 
You are receiving this mail because:
You are on the CC list for the bug.