Uses Valgrind for ExperimentalMemCheck in ctest

[weslleyspereira]

Uses valgrind if it is available.

TO-DO

• Try to optimize the runs with Valgrind. The test is currently taking too much time to complete.

[weslleyspereira]

Related to #551 (comment)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (33b3a45) 0.00% compared to head (7ebfd17) 0.00%.
Report is 2 commits behind head on master.

❗ Current head 7ebfd17 differs from pull request most recent head cfe1114. Consider uploading reports for the commit cfe1114 to get more accurate results

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #844   +/-   ##
=======================================
  Coverage    0.00%    0.00%           
=======================================
  Files        1918     1918           
  Lines      188645   188645           
=======================================
  Misses     188645   188645           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[weslleyspereira]

I believe this is also related to #790.

[weslleyspereira]

Related to #911 and #790.

[weslleyspereira]

Valgrind seems to be working great with LAPACK now. It has found 2 possible memory leaks in example_DGESV_colmajor and example_DGESV_rowmajor. I am taking a look at those.

[langou]

Wow! That's amazing. Thanks @weslleyspereira !

[weslleyspereira]

It should be ready for merge.

There are some spurious defects found in by Valgrind only when used in the Github CI. I am not sure where they come from.

1. no information is provided in the MemoryCheck files.
2. I can't reproduce it locally.

So, I opt for ignoring those defects in the CI for now.