On May 13, 2008, at 2:58 AM, Wen Hao Wang wrote:

I run client/mtt with debug model, and found mtt script exited in following code section of lib/MTT/Module.pm.

32 my $str = "require $module";
33 Debug("Evaluating: $str\n");
34 my $check = eval $str;

When require error occured, $str was set to "MTT::Test::Specify::Simple
Simple".


Note that this *usually* indicates an error in the INI file. Perhaps we should have better error checking/reporting for this case...

Can you send your ini file?

This is not a valid Perl module, so mtt failed. I had no time to probe why/where such value was got. I just modified Module.pm to work around the issue.

[root@host MTT]# !diff
diff primary/Module.pm ./lib/MTT/Module.pm
31a32,38
> if ($module =~ m/Simple/)
> {
> if (length($module) > 26)
> {
> ($module) = $module =~ m/(.*::Simple)/;
> }
> }


This should not be necessary; we all run MTT with the Specify::Simple module every night.

With this workaround, $str was changed back to "MTT::Test::Specify::Simple", and mtt script could continue. Ater all the tests finished, I found some MTT cases failed. For those failed cases, who will probe it? Or how can I probe them? Here list one section that returned error.

| command | mpirun -np 2 --mca btl tcp,self --prefix /LTC/MTT/ installs/Urs4/install |
| | t_win |
| duration | 0 seconds |
| exit_value | 1 |
| result_message | Failed; exit status: 1 |
| result_stdout | WIN_BASIC: passed |
| | WIN_OVERLAP: passed |
| | WIN_OVERLAP2: passed |
| | WIN_OVERLAP3: passed |
| | LOCK_EPOCH: passed |
| | FENCE_EPOCH: failed |
| | MISC: passed |
| | POST_EPOCH: passed |
| | WIN_BASIC: passed |
| | WIN_OVERLAP: passed |
| | WIN_OVERLAP2: passed |
| | WIN_OVERLAP3: passed |
| | LOCK_EPOCH: passed |
| | FENCE_EPOCH: passed |
| | MISC: passed |
| | POST_EPOCH: passed |
| | t_win: FAILED 1 |
| start_timestamp | Tue May 13 04:54:59 2008 |
| test_name | t_win |
| variant | 2 |


These results are reported to the central OMPI MTT database where we get e-mail reports each morning.

This particular failure is a known issue; I believe there's a trac ticket open on it.

--
Jeff Squyres
Cisco Systems

Reply via email to