Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ptp-dev] Mars M6 parallel package

Greg
I created bug 463532 for the problem with the 'Advance' label in the new remote connection dialog.

I created bug 463534 for the local synchronized project problem.

I created bug 463535 for the local connection resources tab problem.

My problem with failing to find ompi_info was a setup problem. Adding /usr/local/openmpi/bin to PATH in my ~/.bashrc fixed that.

Dave
On 03/30/2015 05:26 PM, Greg Watson wrote:

On Mar 28, 2015, at 6:01 PM, David Wootton <drwootton@xxxxxxxxxx> wrote:

I did some testing with this build on Fedora 20 Linux 64 bit and have the following observations. I have not been following PTP development much over the past year, so I may be misunderstanding how some things are supposed to work and don't know how much of what's below is because I did something wrong.
  1. In creating a synchronized project, I opened the new connection dialog. I noticed that near the bottom, the 'Advanced' widget reads 'Advance', not ‘Advanced'
Needs bug.
  1. In creating a synchronized project (shallow) , I defined a new connection Local for host localhost and using password authentication. I specified the remote directory as the directory containing shallow source on my local machine, since I only have this single machine and accepted the default for the project directory. After I clicked Finish a shallow project was created, but contains only an includes directory and no source files. I tried clicking Synchronize->Synchronize All and that did not update the project. I then tried adding filters to include Makefile and *.c and tried to synchronize again but nothing changed.  
Do you have git installed locally? What is your local machine? I think this probably needs a bug.
  1. I tried deleting the shallow synchronized project and in doing so got a warning that 'shallow' contains resources that are not in sync with '/tmp/ws/shallow'. Press 'Continue' to force delete. I clicked Continue and the project was gone, so I guess it works, but I am confused about why I got the message since synchronization didn't seem to be working.
  2. Then I created a 'Local C' project to continue testing
  3. If errors occur resulting in error log entries, I get some sort of an 'alert' asking me if I want to enable error reporting with enable and disable as choices. However, if I try to click either of these options, nothing happens, as if they are not really active. After seeing this a bunch of times, I was finally able to click 'enable'. I'm not sure why. Maybe I got better at hitting the hot spot in the enable label.
This is the new error reporting tool (https://docs.google.com/document/d/14vRLXcgSwy0rEbpJArsR_FftOJW1SjWUAmZuzc2O8YI/pub). I’ve also found that it’s dialogs appear over other dialogs that are modal. You have to move the error dialog away in order to click on the modal dialog. You can disable it under General>Error Reporting
  1. I created a parallel run configuration and selected generic MPICH2 as a target system configuration. If I select local as the connection type, then the resources tab does not have any MPICH2 settings displayed and I cannot run the program without getting errors complaining about incorrect number of processors. 
Sounds like the path is not correct somehow. May be a bug?
  1. If I select remote as the connection type and select my localhost connection, then the resources tab displays the MPICH run parameters. I specified number of processes as 5, as required to run shallow. When I clicked Run I got two error log entries telling me the server connection does not support the required services and the program does not run. My MPICH is version 3.0.4
Sounds like the same bug in you other email.

  1. I deleted the run configuration and recreated it as an OpenMPI run configuration. I had the same problem with the resources tab where when I clicked local, there were no OpenMPI parameters displayed. If I clicked remote then I did see the OpenMPI options, but when I tried to run the program I got an error message telling me that ompi_info could not be found, even though it is supposed to be in my path. My OpenMPI is version 1.6.5.
Sounds like an environment problem.

  1. I looked at preference settings looking for settings that might be specific to MPICH or OpenMPI and didn't find any. When I exited the preferences dialog, I got another popup dialog about a preference recorder from something called Oomph that I don't know anything about asking me if I wanted to record preference tasks for /instance/org.eclipse.ptp.rm.jaxb.control.core/Log File Path. I don't think I clicked anything in preferences, but the Jaxb preferences was the last active page before I closed the preferences dialog.
Ooomph is the new Eclipse installer: https://wiki.eclipse.org/Eclipse_Oomph_Installer
You should complain to them about this :-)
  1. Once I set the path to mpi.h in my project preprocessor include path, CODAN seems to work correctly, where I have no red markers in the right margin, popup text for MPI symbols seems to work, popup test for MPI functions seems to work, although that seemed to work before I set the include path, and cross references seem to work.
  2. MPI artifacts seems to work, MPI barrier analysis seems to work, where it reports 1 error at line 88 and 90 in main.c (I think I have one of the buggy versions of shallow)
  3. The parallel debug and monitoring perspectives open. However, when I opened the monitoring perspective I got an error log entry /tmp/ws/.metadata/.plugins/org.eclipse.ptp.rm.lml.monitor.core/monitors.xml (No such file or directory)
That’s normal for a new workspace.
Thanks for the testing!

Greg


_______________________________________________
ptp-dev mailing list
ptp-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-dev


Back to the top