> I'm using submit-generic, and in the commands I'm submitting,
> I want to be able to use exit(2) to tell rush to requeue the frame
> given certain conditions.
>
> But for some reason, the submit-generic script is mapping all non-zero
> exit codes to cause a 'Fail'. eg:
>
> ###
> ### tahoe.128: 0001
> ###
> --------------- Rush 102.42a7 --------------
> [..]
> -- Command: perl //myserver/jobs/rushscripts/submit-generic.pl -render 1 0 perl
> [..]
> Executing the following in 'perl' (perl):
> $|=1;
> print "This is a test of a retry\n";
> sleep(10);
> exit(2);
>
> --- START
> This is a test of a retry
> --- FAILED: EXITCODE=2
>
> ..and the frame shows up as 'Fail', instead of 'Que'.
Hmm, yes. 6 years and no one's noticed that, not even me!
Looks like the submit-generic script isn't passing your exit code
back to rush, it's short circuiting all non-zero exit codes to 1.
It's a simple fix to the submit-generic.pl script. I'm surprised
I never caught that.
1) Find this block of lines in submit-generic.pl:
---- snip
unless ( close(SHELL) )
{
my $exitcode = ( $? >> 8 );
if ( $exitcode )
{ print STDERR "\n--- FAILED: EXITCODE=$exitcode\n"; exit(1); }
}
---- snip
2) Change those lines to instead read:
---- snip
unless ( close(SHELL) )
{
my $exitcode = ( $? >> 8 );
if ( $exitcode == 2 )
{ print "--- RETRY: EXIT CODE WAS $exitcode\n"; exit(2); }
elsif ( $exitcode != 0 )
{ print "--- FAILED: EXIT CODE WAS $exitcode\n"; exit(1); }
print STDERR "\n--- SUCCEEDED\n";
exit(0);
}
---- snip
I'll include this fix in the next release of Rush: 102.42a8.
|