[torqueusers] Bug? in torque 2.0.0p2

[Garrick Staples]

I've committed your patch, the patch below, and another one for arst
decoding.  Please thoroughly test the next snapshot.

On Tue, Dec 20, 2005 at 04:49:21PM -0800, Garrick Staples alleged:
> On Tue, Dec 20, 2005 at 03:21:05PM -0800, Garrick Staples alleged:
> > On Mon, Dec 19, 2005 at 11:17:12AM +0100, ?ke Sandgren alleged:
> > > On Sat, 2005-12-17 at 11:29 +0100, ?ke Sandgren wrote:
> > > > > This means the child process that will eventually become the job has
> > > > > died.  Unfortunately, this is really really hard to debug.
> > > > > 
> > > > > Have you tried 2.0.0p4?  We're always fixing bugs and have improved
> > > > > logging in this area.  Be sure you configure with --enable-syslog.
> > > > > 
> > > > > One unfixed bug that can cause this is with the job's environmental
> > > > > variables.  Vars with newlines and commas can break things.
> > > > 
> > > > It's still there in 2.0.0p4 although the error message this time says
> > > > Bad file descriptor (9) in TMomFinalizeJob3, read of pipe for sid failed
> > > > for job 200045.ingrid-h.hpc2n.umu.se (0 of 8 bytes)
> > > > 
> > > > I'll try to find this next week...
> > > 
> > > The problem was a newline in one of the env variables.
> > > 
> > > I have written a patch for qsub to handle this and a few other things
> > > regarding the set_job_env function. Take a look at it and use the parts
> > > you feel is worth using.
> > 
> > I tried your patch, and it stops the segfaulting, but things are still
> > not quite right.
> > 
> > Given...
> > BREAKME=$'\n\n,foo=me,blah\n"\n'
> > 
> > In a job, I get...
> > BREAKME=
> > ,foo=me,blah
> 
> I don't know yet if this is breaking something else, but it Looks like
> src/lib/Libattr/attr_fn_arst.c:encode_arst() isn't escaping newlines:
> 
> @@ -361,7 +361,7 @@ int encode_arst(
>  
>    for (pc = attr->at_val.at_arst->as_buf;pc <
> attr->at_val.at_arst->as_next;++pc) 
>      {
> -    if ((*pc == '"') || (*pc == '\'') || (*pc == ',') || (*pc == '\\'))
> +    if ((*pc == '"') || (*pc == '\'') || (*pc == ',') || (*pc == '\\') || (*pc == '\n'))
>        ++j;
>      }
>  
> @@ -384,7 +384,7 @@ int encode_arst(
>  
>    while (pfrom < end) 
>      {
> -    if ((*pfrom == '"') || (*pfrom == '\'') || (*pfrom == ',') || (*pfrom == '\\')) 
> +    if ((*pfrom == '"') || (*pfrom == '\'') || (*pfrom == ',') || (*pfrom == '\\') || (*pfrom == '\n')) 
>        {
>        *pc++ = '\\';
>        *pc   = *pfrom;
> 
> 
> -- 
> Garrick Staples, Linux/HPCC Administrator
> University of Southern California



> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers


-- 
Garrick Staples, Linux/HPCC Administrator
University of Southern California
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20051220/133b18bb/attachment.bin