[haiku] Re: Any plans for pthread support?
- From: Marcus Overhagen <marcusoverhagen@xxxxxxxx>
- To: haiku@xxxxxxxxxxxxx
- Date: Wed, 23 Sep 2009 11:48:03 +0200 (CEST)
Colin Günther <coling@xxxxxx> wrote:
> You can take a look of my implementation here:
> http://dev.osdrawer.net/repositories/changes/haiku-wifi/trunk/wifi/libs/free
> bsd_network2/kern_condvar.c
I guess you want feedback, so let me tell you that the code is way to funny
and contains at least one serious bug.
Instead of testing excessivly single error codes using bloated switch
statements,
it's much better to test for status != B_OK
Your switch cases all miss a "default" for unexpected error codes.
For example, Haiku adds B_NOT_ALLOWED and B_NO_MORE_SEMS
error codes for the semaphore functions which are not documented yet.
You don't need to initialize variables, like
status_t status = ENOERR;
when they get a value assigned as very next step
status = acquire_sem_etc(cvp->sema, 1, B_ABSOLUTE_TIMEOUT,
absoluteTimeout);
This is a bug:
} while (B_INTERRUPTED && (system_time() < absoluteTimeout));
Testing for a single status only, as in
if (status == B_BAD_SEM_ID)
can easily miss other errors.
Last, I wouldn't put misleading information into panic messages, like:
31 panic("%s: B_BAD_VALUE. Wow, either someone altered
my source "
32 "code or a memory error caused a swap in my
hard wired"
33 "bit.\n", __func__);
regards
Marcus
Comedy-Battle! Jetzt ablachen und voten! Neue Comedians, Spontanhumoristen und
Nachwuchs-Spaßmacher machen gute Laune - bestimmen Sie den Wochensieger!
http://comedy-battle.arcor.de/
Other related posts:
