Consider using severity argument to mlint executable

[rlivings39]

Saw the great addition by @RobinTournemenne in https://github.com/Gimly/vscode-matlab/pull/106

There is an undocumented -severity flag to mlint that returns a severity number:

$ mlint -id -severity f.m
L 2 (C 3): NOPRT: ML0: Terminate statement with semicolon to suppress output (in functions).

the ML0. Those are the numbers documented in:

https://stackoverflow.com/questions/35898444/find-category-of-matlab-mlint-warning-id/36058284

Creating this issue in case someone wants to switch to using that. Some caveats:

• -severity is undocumented and unsupported so it will likely change behavior in the future
• -severity was added in R2016b. Using it in a release older than that you'll get an extra error:

$ mlint -id -severity f.m
L 0 (C 0): BDOPT: Option '-severity' is ignored because it is invalid.
L 2 (C 3): NOPRT: Terminate statement with semicolon to suppress output (in functions).

[RobinTournemenne]

Hahaha XD. The time I spent checking every linting message... Well could keep my solution with a if statement as long as Matlab has this -severity flag. I am not even sure it is a good idea because I do think Matlab will keep this -severity flag.

[Gimly]

I think I could add a switch in the config file, like that we can keep both your implementation which will work for all versions, even pre R2016b, and a new one with the flag. This way even if they remove it in a future version we will still have a fallback.

[RobinTournemenne]

Seems good to me.

[otaithleigh]

It appears the older way of doing this is the -m0, -m1, -m2, and -m3 flags, in order of increasing severity. Some additional parsing would have to be done, because -m0 returns (less severe?) warnings and everything from -m1 as well.

Found this article from 2013 while looking for other stuff, and was about to open an issue to bring it your attention: http://undocumentedmatlab.com/articles/parsing-mlint-code-analyzer-output The article is about parsing output within MATLAB, but all the options I've tried so far work with the command line version of mlint.