Use std::shared_ptr instead of our own RC* function. #399
Conversation
…l, use latest mkl.
Codecov Report
@@ Coverage Diff @@
## master #399 +/- ##
==========================================
- Coverage 70.09% 70.07% -0.03%
==========================================
Files 96 95 -1
Lines 6604 6576 -28
==========================================
- Hits 4629 4608 -21
+ Misses 1975 1968 -7
Continue to review full report at Codecov.
|
|
It looks like this is an heisenbug in travis. 2 failures previously (linux), one now (mac which was passing previously, and the change is a dummy one - removal of a comment -). So I think this is ready to be merged... Please do not squash the merge, as it makes further merges difficult (this was done with the previous PR, which obliged me to do a manual merge with all the time and the risks this incurs). |
|
@papadop you've been missing my comments or just ignoring them but I am not ok with adding a dependency on boost and I don't want to merge code that makes CIs red. sorry for being opinionated. I find this other PR #397 more problematic and urgent as it reveals an actual bug that prevents me from moving forward with mne-openmeeg integration |
|
Sorry but I have not seen your comment... not on this PR at least. As for the red, read my comment. It is a problem with travis, not with the code. Two successive runs (with no change) result in two different results... If travis is not reliable, then we will have to re-consider this policy.... Hopefully it is just a temporary problem. |
|
As for PR #397, I'm on it... But the solution is more complex than you may think. The problem is hidden in the mess of our mesh/geometry structure and a probable solution is .... shared_pointer !!! I spent most of my week-end in trying to clean the mesh/geometry code to the point it will be more manageable.... Yet another can of worms I opened !!! So I'm on it, but it evolves as fast as I have time to deal with it. |
|
Travis is green ;-) |
| # CHECK_CXX_FEATURE(HAVE_MATH_FN_IN_NAMESPACE_STD math_fn_in_namespace_std.cpp "has C math functions in namespace std") | ||
| # CHECK_CXX_FEATURE(HAVE_MATH_ABSINT_IN_NAMESPACE_STD math_absint_in_namespace_std.cpp "has C math abs(integer type) in namespace std") | ||
| # CHECK_CXX_FEATURE(HAVE_COMPLEX_MATH_IN_NAMESPACE_STD complex_math_in_namespace_std.cpp "supports complex math functions are in namespace std") | ||
| # CHECK_CXX_FEATURE(HAVE_ISNAN_IN_NAMESPACE_STD isnan_in_namespace_std.cpp "has isnan function in namespace std") |
There was a problem hiding this comment.
FYI @papadop these had been commented out as it was not used and it was increasing cmake bulid time for nothing.
There was a problem hiding this comment.
Yes pre-approved. Maybe you can integrate it with your old CMake removal. I already have merged this branch....
Replaces PR #396 (that I have not been able to rebase)....
This PR removes the use of our own shared pointer solution in favor of the standard std::shared_ptr. Hopefully it will also help to improve the python binding as the std::shared_ptr is already handled by swig. This also remove some files which we no longer have to maintain and simplifies a little bit the code. Tested on fedora 31, let's see what the other platforms will say....