New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
-DBUILD_PARALLEL=ON fails on mac (at least) #667
Comments
I don't see gen_pngs.py in the repo and none of the tests fail in this way. Do you have a script that can reproduce this? |
@dlonie If you can source the runtime script and run regular tests, then probably look into the setup_runtime.sh and see if there is something missing for Mac. |
The only OS-specific bit in there is:
which shouldn't be causing this sort of problem. I'll need someway to reproduce this to figure out what's going on. |
If that's the case, I think the issue is this then: ImportError: dlopen(/Users/mccoy20/uvcdat-app/14-09-16/Externals/lib/paraview-4.1/site-packages/vtk/vtkIONetCDFPython.so, 2): Symbol not found: __ZNSt8ios_base4InitD1Ev Were you able to run tests with the paraview build? Can you post results on the cdash? |
Looking good. I think the other tests are failing because the ParaView does not has your change to CellToPointData fix. Is that merged into the master yet? Thanks |
@doutriaux1 Can you reproduce this bug anymore? |
was it an issue on the vtk/paraview repos? should I try from master again? |
I can't reproduce it, so I'm not sure what's going on -- the test script used in the report isn't in the repository. Is there another way to trigger this using a vcs script? |
@dlonie I think the real issue is that -DCDAT_BUILD_PARALLEL=ON breaks the build, try to build with this option, source setup_runtime and import vcs and plot something, it should fail |
I set that on, rebuilt in a clean directory, and ran the tests. I didn't see this error (some image tests failed, but it looks like it's because the vtk/uvcdat-master and paraview have diverged.) This is on linux, though...I don't have a mac, so if it's mac-specific I'll have to find one to test on. Maybe @aashish24 has a mac I can use....? |
we experienced it on both mac and linux, let me test again on both. |
@doutriaux1 @dlonie ran the entire test suite you can find it here: http://open.cdash.org/viewTest.php?onlyfailed&buildid=3514542 |
I am unable to produce it on my Mac and Linux as of last week. |
just failed on my ubuntu |
@doutriaux1 we are unable to reproduce this. The dashborad (salix) seems to be fine. http://open.cdash.org/buildSummary.php?buildid=3517849 How we can reproduce this? Can you setup a dashboard on this machine? or Submit ctest -D Experimental |
|
That's not reproducing on mine, David's and most importanly on Salix dashboard (all Linux) Can you submit resutls
Or setup a dashboard on this machine please |
seems to work, it only the CDAT_BUILD_PARALLEL=ON alone. that seems to fail |
That's what @dlonie tried. @dlonie can you confirm? I can run just with that too. But I think it will be most beneficial, if you can post your resutls. Without that it will be hard to see what's going on. ctest -D Experimental Also, please upload or your CMakeCache.txt |
Trying now... |
Built and ran tests on Arch Linux. Some dashboard, but about usual for me: http://open.cdash.org/viewTest.php?onlyfailed&buildid=3518414 Most are ESMF linking problems that have always been there on my system, the rest seem to be rendering issues since VTK/ParaView have diverged for the uvcdat branches. |
Yup that is expected unless we update ParaView (which we don't today) |
@doutriaux1 is this still an issue? |
I believe so. But let me try it again this morning. |
nope still here on my mac 10.8 |
here's what I get:
|
Also mentioned here: #789. For one, I would recommend trying with It looks to me like IceT is using |
ok will try the make -j1, but if that works then we need to figure out which dependencies are wrong. |
make -j1 output
|
Looks like ESMF is broken. Can you confirm? If that' the case, we need to ping ESMF folks. |
it's not ESMF because -CDAT_BUILD_MODE=ALL works and that turns on parallel. So it's definitely our build. |
So basically in PARALLEL ON mode it seems like ParaView and ESMF both are failing. |
The following worked for me at commit b7aada3 OS X 10.10.2 with
Running ctest now, but they appear to be passing. |
yes but only in parallel=on build_mode=all works. must be dependencies order or missing |
@jbeezley i will try with these options |
Charles, something is missing. The issue label says that we have problem when parallel=on fails to build. If the issue is something else, I would suggest we create a new issue, provide the exact command that you used and then can figure out what is wrong. |
well that's the issue no? -DCDAT_BUILD_PARRALEL=ON fails. Feel free to create another issue and close this one. |
Oh .. what you said confused me: "yes but only in parallel=on build_mode=all works. must be dependencies order or missing" |
sorry, I forgot the "dot" between the two options. BUILD_MODE=ALL does work (and it triggers PARALLEL=ON) so I think it is a dependencies order or dependencies missing. |
need to try on 10.9 too. Since it seems to work on 10.10 for @jbeezley |
@doutriaux1 can you try please? |
@aashish24 @jbeezley I get the exact same thing on 10.9
|
Closing, we don't have ParaView in the superbuild anymore. |
maybe as simple as a tweak to setup_runtime.[c]sh
The text was updated successfully, but these errors were encountered: