New feature Mex Out-of-Process

In the R2019a, there is a new feature that I think it's useful for people like me(us) that uses heavily MEX in the projects.

https://fr.mathworks.com/help/matlab/matlab_external/out-of-process-execution-of-c-mex-functions.html

The problem is it seems support only C++, and not C and Fortran.

My question is why and more importantly : is it in the roadmap of TMW that MATLAB will eventually supports this feature for C/fortran or TMW force users to migrate their MEX file to C++ to be able to use the external function in out-of-process manner?

[James Tursa]

Have you tried to feed mexhost a regular C-mex function?

[Bruno Luong]

Yes it returns an error

>> feval(mh,'myCmex',17)
Invalid C++ MEX-file 'myCmex'

[Bruno Luong]

Any authority can please shed a light?

[Jan]

@Bruno: It happens rarely, that authorities of the MathWorks staff reveal the details in the forum, especially if they concern features of future releases.

[A.B.]

This is a highly useful needed feature in C/Fortran as well. Not all programmers code in C++. I hope MATLAB takes this issue up soon.

Answer 1

[Steven Lord]

Jan is correct that generally we don't discuss whether or not something is in our future plans.

Bruno Luong, A King, if you want to request that we add C and/or Fortran support to the out-of-process MEX feature please let Technical Support know via the telephone icon in the upper-right corner of this page. Including details about how you would use this feature if it were available and what benefit you would hope / expect to gain in your enhancement request would be useful information for the developers to consider when reviewing the request. "I want feature X because it would let me do Y." can be a stronger motivator than just "I want feature X."

[A.B.]

Steve, Thank you. I will try to explain the reasoning in a ticket submission. But I do not understand the need for such explanation. It is needed in C/Fortran for the same reasons it was originally deemed necessary in C++. Just because of this feature not being available in C/Fortran I had to rewrite and recompile several sections of a large package so that upon premature ending of the C/Fortran codes, the entire MATLAB does not crash. This problem is unique to MATLAB, and no other language. But just because of this missing feature in MATLAB, the code had to be redesgined to handle such unexpected errors.

[Bruno Luong]

Debugging MEX file is not convenient if it's used the same process with MATLAB.

If user see a pointer wrongly set during a debugging, if he/she stop prematuraly the MEX, it crashes, if he/she continues it crashs. This behavior is for C/Fortran/C++. Then she/he has to start the whole thing over again.

So it nice to have this feature on ALL external language that MATLAB supports.

It's just plain odd that such feature applies only for C++.

[Joseph Becker]

I do have >20,000 lines of legacy FORTRAN that took over a year to get sort of running in a MEX because the debugging was so insanely difficult to do without something like GDB connecting to process outside of MATLAB. I'm sort of mystified why this out-of-process for C/Fortran is difficult to add, maybe it was just an oversight but let me assure that few things are as difficult to debug as a MEX. And given Matlab crashes at the drop of a hat when the MEX FORTRAN has an issue it's just about impossible.

Hmmm, guess I'm an idiot, but I don't see "the telephone icon in the upper-right corner of this page".