Error "Invalid MEX-file" in unstable version

[Stefan_Boeters]

Some days ago I've started working with the latest issue of the unstable version: "master-2017-05-12-ae50f25"

Now, in a model run that was successful under version 4.4.3, I get the following error message about an "invalid MEX-file". What is going on here?

Invalid MEX-file 'M:\p_HerzSaffier\Dynare\4.5\matlab\..\mex\matlab\win64-7.8-9.2\mjdgges.mexw64': Error performing inpage
operation.

Error in dyn_first_order_solver (line 216)
[err, ss, tt, w, sdim, dr.eigval, info1] = mjdgges(E, D, DynareOptions.qz_criterium, DynareOptions.qz_zero_threshold);

Error in stochastic_solvers (line 264)
[dr,info] = dyn_first_order_solver(jacobia_,M_,dr,options_,task);

Error in resol (line 144)
[dr,info] = stochastic_solvers(dr,check_flag,M,options,oo);

Error in check (line 73)
[dr,info,M,options,oo] = resol(1,M,options,oo);

Error in estimate_v6_7_combi (line 1488)
oo_.dr.eigval = check(M_,options_,oo_)

[jpfeifer]

Which operating system?

[Stefan_Boeters]

Operating system is Windows 7.

[jpfeifer]

I googled the error message and it mostly turns up problems related to the harddisk: http://www.tech-faq.com/error-performing-inpage-operation.html
I am not sure this is a Dynare problem.

[Stefan_Boeters]

Part of the problem is that I have two instances of Matlab open. I now noticed that in one of the instances, everything runs fine, whereas in the other instance (even if the first is not busy currently), I get the "Invalid MEX-file" message.

This seems related to Dynare at least in so far that everything runs fine (in both instances of Matlab) with Dynare 4.4.3.

[jpfeifer]

Strange. Which Matlab version are you using? And does the problem occur in a predictable way? Or does it disappear after a reboot.

[Stefan_Boeters]

Matlab R2017a. -- I'll keep experimenting and will let you know whether the problem disappears or a systematic pattern emerges.

This may take some days. At the moment I'm happy that one of the Matlab-instances does the job.

[jpfeifer]

It may have to do with Matlab's new JIT compiler, which complicates using routines in simultaneously running Matlab sessions

[Stefan_Boeters]

After restarting everything, the error does not occur any more.

Thanks for having a look into the issue.