Inconsistent ctfroot() output

Does output of ctfroot depend on any machine parameters? Is it expected to be consistent across installations?

I had created an appdesigner app for deployment and had been maintaining it. The app requires a mat file and some directories to run and these are automatically created if the application fails to detect them upon launch (the same thing happens on first launch). The required files are packaged into the compiled app. Until now it worked well upon installation. Last week I was told it's throwing an error upon first launch on a new installation and quitting. I investigated and found that file-copying op (from packaged cache to working directories) was failing citing "not found" error. Upon closer look the output returned by ctfroot was "C:\Users\User\AppData\Local\Temp\User\mcrCache23.2\App_60\" on this machine. Where as the correct path would be "C:\Users\User\AppData\Local\Temp\User\mcrCache23.2\App_60\App\". The latter path is what ctfroot() returned so far and the former (without the /App/) is what I get now.

I have some PCs on which the app worked with the same installer. I did update the application but the updates have nothing to do with the problem. The MCR version is the same on the machines this app was installed on a couple of months ago when it worked and the machine this was installed last week where it has this problem. I even tried uninstalling the app and MCR and reinstall using the older version of installer (with embedded MCR) but the problem remains. Any idea what's different now or on this machine? Is ctfroot expected to be differnt across installations?

[Walter Roberson]

Generally speaking, ctfroot() needs to be adaptable to the directory that the app was installed into. That directory would have to change if, for example, MCR was installed into a different drive. (There are some systems that have no C drive.)

This does not explain the different paths you are encountering.

[cr]

I totally agree and what you say is kinda expected. On all machines the app is installed on, the installation directory has been the same -the default.

[cr]

I found the same problem being reported here but no clear answers.

https://in.mathworks.com/matlabcentral/answers/448677-why-does-my-ctfroot-have-unexpected-subfolders

[Adam]

Is this definitely the same installer on two different computers or different installers?

I found an annoying issue previously in that my current directory at the time of building the app changes the folder structure where it gets deployed on the target machine. This wasn't what I expected, so now I always make sure I navigate to the same folder when I open the Compiler to build my executables so that the relative paths will be the same.

I don't know if this is the issue you have, it sounds different, if it's the same installer, but outside of that I've not had issues with ctfroot until R2025a where everything is just such a mess I had to drop back to using R2024a for my compiling of apps, because it wouldn't find my dlls etc despite them being put in the right place, suppoedly, which was also some kind of relative path issue maybe. I don't know which version of Matlab you are using though, since the MCR cahche folder name is always different to the Matlab version.

[cr]

I'm using R2023b. I definitley tried the same old installer that on other machines worked fine. I can tell this from the timestamp on the installer. I too suspected that the current directory at the time of compilation was the culprit, so I restarted matlab and recompiled with the current directory restored but it didn't make a difference.