Re: timeout fix for linux

The man timeout page says:
"timeout's exit status is the exit status of the specified command
or 1 in case of a usage error."

My problem was that the executed command failed with exit value other than 0 (eg. sqlplus 
and 174) in time. But timeout returned 0, so the script assumed the call was successful. I 
only touched the code of the returned value surrounding status with the appropriate macro.

Sure, it may interfere when the command also returns -1 or 255 but these values usually 
means fatal, unknown errors without useful meanings. I think this is not a problem (at 
least the problem remains problem), while reporting success instead of failure is danger.

That's all. I thought others may also assume the above operation.. Feel free to include or 
drop.

BR,
Laszlo

Wietse Venema wrote:
> Keresztfalvi Laszlo:
>> Hello,
>>
>> since I did not find other address to send a patch, here it is:
>>
>> in src/misc/timeout.c (at least on Linux):
> 

> This patch does not correctly report that the process was killed
> by a signal.
> 
> 	Wietse
> 

>> #include <sys/wait.h> and returning WEXITSTATUS(status) is required to correctly return the executed command's 
>> exit value.
>>
>> Best regards,
>> Laszlo
> 
> [ text/x-patch is unsupported, treating like TEXT/PLAIN ]
> 
>> diff -ru tct-1.16/src/misc/timeout.c tct-1.16-fixed/src/misc/timeout.c
>> --- tct-1.16/src/misc/timeout.c	2004-10-18 16:59:31.000000000 +0200
>> +++ tct-1.16-fixed/src/misc/timeout.c	2007-01-19 11:03:58.000000000 +0100
>> @@ -34,6 +34,7 @@
>>  /* System libraries. */
>>  
>>  #include <sys/types.h>
>> +#include <sys/wait.h>
>>  #include <signal.h>
>>  #include <stdlib.h>
>>  #include <unistd.h>
>> @@ -107,6 +108,6 @@
>>  	alarm(time_to_run);
>>  	while ((pid = wait(&status)) != -1 && pid != child_pid)
>>  	     /* void */ ;
>> -	return (pid == child_pid ? status : -1);
>> +	return (pid == child_pid ? WEXITSTATUS(status) : -1);
>>      }
>>  }
>