[Win] R2015b crashes often, is more unstable

Has anyone noticed that Matlab has become more unstable under R2015b? In my use case I run some intensive data crunching and especially when trying to abort with CTRL+C, the system often crashes. I have done that countless times with the R2014b release and almost never had a problem.

I get the following crash details:

Problem signature:
  Problem Event Name:  APPCRASH
  Application Name:  MATLAB.exe
  Application Version:  8.6.0.0
  Application Timestamp:  55bacb01
  Fault Module Name:  m_lxe.dll
  Fault Module Version:  0.0.0.0
  Fault Module Timestamp:  55badb54
  Exception Code:  c0000005
  Exception Offset:  0000000000023477
  OS Version:  6.1.7601.2.1.0.256.4
  Locale ID:  2057
  Additional Information 1:  ac3e
  Additional Information 2:  ac3e992fd57a22b183ed49bb92ee620c
  Additional Information 3:  c613
  Additional Information 4:  c613c727341c18b2b2285b7aeda350a6

That new engine, can't deal with my ... methods

UPDATE 01 Feb 2016

The crash does NOT produce dump files, but I managed to reproduce the situation that generates the issue.

If I run the following code and I CTRL+C (you might need to try a few times before it crashes), then I get the APPCRASH:

for ii = 1:1e4
    load('T00001.mst','-mat');
    if mst.To % Problem is linked to this
    end
end

The code loads a table of name mst and then checks that the column of the table meets some condition (I simplified down the code to the bare minimum). If I remove the condition with the dot indexing CTRL+C won't crash Matlab. Hence, it seems something wrong goes with the interaction of load and table indexing. The condition alone, on a previously loaded table, does not crash Matlab.

[Walter Roberson]

You have ... methods? Is that Evil Villain Laughter #7 that I hear in the background?

;-)

[Yuebin Zhou]

You can post some crash dump files. That would be helpful.

You can find crash dump files in the folder returned by the following command:

>>winopen(tempdir)

Answer 1

[Oleg Komarov]

Support suggested to try the pre-release R2016a and I did not manage to crash Matlab on that. You can infer the rest.