My original post :
>
> Dear Managers,
>
> I have few problems (twice a year) with the semop system call.
> A program which usually works fine failed on a semaphore operation.
> The errno error code was either no set (0) or set to a strange value
> (not documented). As state in the semop manual page, the program
> considers that the semop operation succeeds if it returns a zero value.
> But I wonder if the semop system call may return a postive value on
> success or if there is a bug in the semop implementation.
> I would be pleased if one could send me the kernel source code of the
> semop system call. Up to now, I have no response from Digital support.
> (I am using DU 3.2C, but I had the same problem on Ultrix 4.3.1, but
> never on AIX, HPUX, Solaris, LynxOS ... this is why I consider that
> my program is correctly written)
>
> Thanks a lot.
>
> Laurent.
>
No response yet.
I have reproduced this problem this week, and hoppefully, I have
added some useful tracing which shows me that all parameters were
correct and that the semaphore id was really in use (ipcs).
In that case, the semop system call returned -1 and errno was set to
ECONNRESET (54) !!
So, there is really a bug in the IPC sempahore implementation.
Laurent.
--
Laurent DENIEL | E-mail: deniel_at_worldnet.fr
Paris, FRANCE | WWW : http://www.worldnet.net/~deniel
All above opinions are personal, unless stated otherwise.
Received on Fri Nov 08 1996 - 19:38:01 NZDT