Fix out-of-bounds bug in mex code

[mducle]

There was an ancient bug in the parallelised eig and chol code which failed to check the number of output variables nlhs and tries to access a pointer which does not exist.

In older versions of Matlab this seems to work (the routines get a pointer to the actual data using mxGetData given the address of an mxArray but never uses that pointer), but with Matlab 2023a and newer this causes a segmentation fault because the internal behaviour of mxGetData / the internal definition of the opaque mxArray class/struct has changed.

Specifically, the code tries to access the output pointers array plhs[] which has nlhs elements long but the code always assumes that nlhs is at lease 2. Hence when nlhs=1, plhs[1] is a junk pointer. I suspect that in older versions of Matlab, mxGetData just returns the input pointer plus an offset, which will work with a junk input pointer as long as no processing is done on the (junk) output. I think that in newer versions of Matlab, mxGetData actually tries to dereference the mxArray structure to access its pr member (the data pointer), and this will give a segfault because the input pointer is junk.

The PR now checks what nlhs is and guards against this out-of-bounds error.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 43.29%. Comparing base (7e87dee) to head (a2bf25c).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #214   +/-   ##
=======================================
  Coverage   43.29%   43.29%           
=======================================
  Files         243      243           
  Lines       16383    16383           
=======================================
  Hits         7093     7093           
  Misses       9290     9290           

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

[github-actions]

Test Results

    4 files  ±0    124 suites  ±0   4m 59s ⏱️ - 1m 32s
  831 tests ±0    813 ✅ ±0  18 💤 ±0  0 ❌ ±0 
2 324 runs  ±0  2 288 ✅ ±0  36 💤 ±0  0 ❌ ±0 

Results for commit a2bf25c. ± Comparison against base commit 7e87dee.

♻️ This comment has been updated with latest results.

[RichardWaiteSTFC]

Works, thanks for quick fit. Checked on IDAaaS using MATLAB 2024a