MATLAB 2020a and prior versions generate pcode warning

I help run a centralized computing site with MATLAB for our users, and at least two versions (2019a and 2020a) complain that the P-file for ver.p is older than the ver.m that's installed and might need to be rebuilt. The ver function seems to be needed by a several functions to do something (check the license?), so the warning message comes up frequently, but it's not an error and doesn't cause code failures, just confusion and trouble tickets from our users. My assumption was that these pcode files were created by The Mathworks and shipped with MATLAB and end users need not worry about them. Any advice? Here's a sample of the message, lightly edited to shorten the paths:

Warning: P-file /xxxxxxx/matlab/2020a/toolbox/matlab/general/ver.p is older
than MATLAB file /xxxxxxx/matlab/2020a/toolbox/matlab/general/ver.m.
/xxxxxxx/matlab/2020a/toolbox/matlab/general/ver.p may be obsolete and may
need to be regenerated.
Type "help pcode" for information about generating P-files.

Users sometimes try to regenerate the P-files themselves, which can lead to some really interesting trouble tickets since they don't have write permissions to the MATLAB installation directory. For the 2020a files, they both have UNIX modify and change dates a few seconds apart on what I presume to be the install date (I'm not the one that did the install).

Answer 1

[Fangjun Jiang]

Someone must have modified and saved the ver.m file by mistake. ver.m and ver.p are built-in files and shouldn't be modified. Restore the original copy for those two files and the problem should be resolved.

The "Date Modified" for ver.m in R2020a is 3/6/2019. For ver.p, it is 1/29/2020.

[Bill Barth]

That seems unlikely since my 2020a files have April 2020 dates on them (which likely go along with our install date, which seems especially likely since their modify and change times are 2 seconds apart, perhaps indicating a copy and a very fast build system) and diff says that our 2020a ver.m matches our 2019b ver.m. Trying to get further apart in time (even within a year like 2019) shows differences (2019a/b ver.m have a big diff). ver.p isn't even among the most recent files in that directory.

Thus the reason I'm here. We could loop over the provided P-files and find the out of date ones and reconstruct them, but that seems like a waste of time if someone has a better idea.

[Fangjun Jiang]

check the date "modified", not "created".

[Bill Barth]

They're the same with stat to all provided digits.

[Fangjun Jiang]

run "which -all ver" to see if there are multiple files.

run "pcode ver.m" and over-write the ver.p file. Make sure nobody else would do the same or have the right to over-write the ver.m file.

[Bill Barth]

No extra P-file or M-file, but it notices the M-file that shadows it:

>> which -all ver
/xxxxxxx/matlab/2020a/toolbox/matlab/general/ver.p
/xxxxxxx/matlab/2020a/toolbox/matlab/general/ver.m  % Shadowed

But we knew there must be an M-file matching our P-file, thus the complaint. I don't think the admin who installed this and the prior version accidentally or purposefully overwrote the P-file.

Also, I should note that while 2019a is fine, 2018b has the same problem and the same complaint about fullfile.m and .p. 2019b, only has the issue with ver.m and .p.

Given that the whole Linux world isn't complaining about this, I agree that this is probably us and not The Mathworks, but I don't think we're doing it on purpose. There must be something systematic about the issue. There are roughly 10k M-files installed in the core MATLAB toolbox, and only 1500-ish P-files. I doubt those are created at install-time, but instead created at the mothership and installed with the product.

I know I could run pcode on ver.m and get rid of the ver.p file I have and not have to deal with the trouble tickets, but I'd like to not destroy the evidence in the process.

[Fangjun Jiang]

Many .m files are for help document, for example abs.m, there won't be one-to-one match of .p file for every .m file.

Anyway, the warning message is annoying. Contact the Mathworks Tech Support. I bet they can tell you what should be the exact "time stamp" on the files and you could eventually figure out the cause and solution.