[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
> P.S. Just curious, but what was the reason you guys are using signals
> instead of threads? To quote Vincent: Not pushing, just asking :-)
:)
I used to have read errors for datafile stuff (maybe write errors
too, can't remember, but very definitely read errors). They were
happening quite seldom, and, IIRC, always at different places.
Switching to SIGALRM fixed it. I never had such problems afterwards.
Some code got patched to handle interrupted IO syscalls, but I
can't recall whether this happened before or after. This was back
when Shawn was still the maintainer.
Which reminds me I'm not actually using SIGALRM, but SIGVTALRM,
as using the former prevents profiling. Don't ask me why, since
there seems to be a dedicated signal (SIGPROF IIRC, but my memory
is fuzzy). It just happens that using SIGVTALRM does the trick.
There does not seem to be adverse effects with timers or VT
switching that I can see, though this is typically stuff that is
kind of hard to spot reliably.
Also, pthreads does not play nice with my GDB. I have an old system
though, that probably has something to do with it. Though Annie
uses a newer one (2.4.13, glibc 2.2 (I think)) and can't get any
profiling done. But that's not Allegro related, *nothing* can get
profiled (except call counts, of course).
--
Vincent Penquerc'h