Confusing warning during compilation: Warning: MATLAB preference entry "SMTP_Username" has been removed.

I use sendmail in a compiled program that sends email via a generic gmail address. Username and password are set via:

setpref('Internet','SMTP_Username',sender)
setpref('Internet','SMTP_Password',passwd)

When compiling with Compiler version: 8.1 (R2020b), I get the following warnings:

Warning: MATLAB preference entry "SMTP_Username" has been removed. 
Warning: MATLAB preference entry "SMTP_Password" has been removed. 

Still the code perfoms OK.

I don't really understand what this means, whether this is a feature that will be removed in future releases...

I could not find any information at mathworks or the internet in general. Any pointer is appreciated!

[Walter Roberson]

odd https://www.mathworks.com/help/matlab/ref/sendmail.html documents it.

[Clayton Gotberg]

My interpretation of the error is that when you run the compiler, MATLAB removes its memory of the listed preferences. Maybe try checking your preference list before and after you run the compiler to see if those values have changed.

setpref('Internet','SMTP_Username',sender)
setpref('Internet','SMTP_Password',passwd)

getpref('Internet')

%% Run compiler %%

getpref('Internet')

[Walter Roberson]

Oh... I wonder...

It seems likely that for some preferences, you would want to take the preference setting of the user who is doing the compiling, and "hard code" those values into the executable, but that for some other preferences, you would want to not want to hard-code the values and would instead want to use the values set by the user at run-time . Perhaps that is what the message is talking about?

For example in this case each user of the executable might have to set their own Environment Variables for SMTP_Username and SMTP_Password, since that is something that is likely to be different for each user.

[Clayton Gotberg]

I wonder if that is either to warn about or to repair this security issue? It seems that you're correct about MATLAB not wanting those preferences to be hard-coded!

[Eric Delgado]

Hi @Ralf, it seems like a "default warning". I am still learning "Matlab Compiler" and made I simple app with no SMTP settings... but I receive this warning anyway.

Let me ask you something... :)

Do you know what is the best approach to deal with a big app that calls Python, for example? Should I build an app just to configure pyenv and my app PATHS, for example?

Answer 1

[Mohammad Sami]

This is the correct behaviour for security reasons. Otherwise this information will be included from the previously set values in your Matlab environment if any. This would have happened regardless of whether you were using the mail function in your compiled program. These values would then have been exposed (visible in plain text) once you distribute your program to others.

Therefore you will need to use the setpref function in your code to set these values before you try to send the emails. You can also prompt the user for these values, then call the setpref function using the entered values.

[Augustin Thepaut]

I actually have the same issue but I do not know what "use the setpref function in your code to set these values before you try to send the emails" means?

Is ther a place where I can run this automatically?

Than you for your help.

[Walter Roberson]

@Augustin Thepaut

Refer back to @Clayton Gotberg comment to the Question, https://www.mathworks.com/matlabcentral/answers/813140-confusing-warning-during-compilation-warning-matlab-preference-entry-smtp_username-has-been-remo#comment_1483060

Your deployed code needs to somehow decide what username and password to use, and setpref() them.

Deployed code will be encrypted if you use MATLAB Compiler, but it would still generally be considered not a good idea to hard-code username and password (for one thing what would you do if you need to change the password?)

So you would probably be wanting to do something like check environment variables (though the security for that is not that great) or prompting the user to input the information every time, or prompt only once but store it (encrypted) in a file.... Once you have obtained the information (somehow) then setpref() as needed.

No, there is no "automatic" way of doing this. Or rather the "automatic" way of doing it is deliberately disabled because of the security risks, and what you decide to replace the automatic way with depends upon your instituational security policies.

[Augustin Thepaut]

Hey! Thanks for the response is such short notice! Ok I understand. However, I am wondering for instance how compagnies send email automatically to their users (for a report for instance), if you cant set this properly?

Not sure I am really clear here but it is just curiousity anyway (more like culture).

Thanks again!

[Walter Roberson]

@Augustin Thepaut

Normally, at the time you compile an executable using MATLAB Compiler, MATLAB Compiler reads the current user's environment variables and turns them into internal setpref() in the executable. However, if the user who is doing the compiling has either of the environment variables SMTP_Username or SMTP_Password set, then MATLAB Compiler issues a warning message and does not insert those values into the executable.

Then at run time, if the code has not been set up to create those environment variables but a user identity is needed, then at runtime the code would look at the SMTP_Username or SMTP_Password environment variables of the user who is running the code.

It is uncommon that you would want the mail username and password of the person doing the compiling to be coded into the executable. In the cases where the person building the code does want that, then there are a number of ways that the code could be written to hard-code the username and password; the exact means used would depend upon the security policies of the organization.

Answer 2

[Derek Wolfe]

I ran into the same issue. My compile script uses the following to clear Internet preferences before building:

% clear email preferences

if (isempty(getpref('Internet')) == false)

rmpref('Internet');

end