Project

General

Profile

Task #1309

Jenkins verification improvements

Added by Teemu Murtola over 6 years ago. Updated over 2 years ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
Jenkins
Target version:
-
Close

Description

Two suggestions for improving the checking done in Jenkins, based on recent experiences (see the commit messages of the linked changes in Gerrit):
  1. Remove the rule that ignores Doxygen warnings from legacyheaders/ (after https://gerrit.gromacs.org/#/c/2516/ is merged). These headers should have been warning-free for the most part of the past year.
  2. Make the main build unstable if it produces CMake warnings (after https://gerrit.gromacs.org/#/c/2514/ or the alternative proposed there is merged). Don't know if some of the current builds produce some other warnings, but if they do, then those should be fixed as well by changing the CMake invocation and/or the CMake code.

History

#1 Updated by Roland Schulz over 6 years ago

1. done.

2. We currently have several warnings in different configurations:
- The fftw library found is compiled without SIMD support
- Please manually add compiler flags to CUDA_NVCC_FLAGS.
- Manually-specified variables were not used by the project: CUDA_NVCC_HOST_COMPILER
- To use GPU acceleration efficiently, mdrun requires OpenMP multi-threading.
- Shared system libraries requested, and static Gromacs libraries requested.

I added a Cmake error parser. At test run with it is build: http://jenkins.gromacs.org/job/Gromacs_Gerrit_master/2111/
If we get rid of those warnings we could enable it.

#2 Updated by Rossen Apostolov over 5 years ago

Can we close this report?

#3 Updated by Roland Schulz over 5 years ago

No. We still have these cmake warnings. If we want to enable the Cmake error parser we need to still fix all warning (e.g. by installing a correct version of fftw) or have some way to distinguish expected warnings from unexpected ones.

#4 Updated by Szilárd Páll over 2 years ago

Roland Schulz wrote:

No. We still have these cmake warnings. If we want to enable the Cmake error parser we need to still fix all warning (e.g. by installing a correct version of fftw) or have some way to distinguish expected warnings from unexpected ones.

This is still not solved, right?

#5 Updated by Roland Schulz over 2 years ago

It seems that at least for Matrix_PreSubmit we don't have any cmake warnings. I enabled the parser. Test build is here: http://jenkins.gromacs.org/job/Matrix_PreSubmit_master/35/
Not sure whether we want to enable it for other jobs.

BTW: I noticed that http://jenkins.gromacs.org/job/Matrix_PreSubmit_master/34/OPTIONS=gcc-5%20gpu%20cuda-8.0%20openmp%20release%20host=bs_nix1204,label=bs_nix1204/consoleFull has nvcc warnings. Would it be valuable to have the nvcc warnings also flag a build as unstable? If so we should have a parser for that and avoid this warning.

#6 Updated by Szilárd Páll over 2 years ago

Roland Schulz wrote:

It seems that at least for Matrix_PreSubmit we don't have any cmake warnings. I enabled the parser. Test build is here: http://jenkins.gromacs.org/job/Matrix_PreSubmit_master/35/
Not sure whether we want to enable it for other jobs.

It think is pre-submit we could enable it. After all, pre-submit should reflect the "most supported" configs, so it'd make sense if those were cmake warning-free.

BTW: I noticed that http://jenkins.gromacs.org/job/Matrix_PreSubmit_master/34/OPTIONS=gcc-5%20gpu%20cuda-8.0%20openmp%20release%20host=bs_nix1204,label=bs_nix1204/consoleFull has nvcc warnings. Would it be valuable to have the nvcc warnings also flag a build as unstable? If so we should have a parser for that and avoid this warning.

See #2141, I've just not gotten to doing it.

Also available in: Atom PDF