CODEWITHSUNDEEP
debuggingmatlaberror-handling
gerrit
gerrit

Reputation: 26465

Debugging back to original error with conditional exception handling, i.e. rethrow

In Matlab, there is, as far as I know, no good way to conditionally catch exceptions (correct me if I'm wrong). The only way is to catch the exception, check the identifier, and rethrow the error if this particular error can not be handled. That's acceptable though inconvenient. However, when I use Matlabs dbstop if error, I end up at the ME.rethrow() line. I'm then unable to dbup back to the place where the original error was caused.

function test_excc

try
    sub_test()
catch ME
    if strcmp(ME.identifier, 'test:notsobad')
        fprintf(1, 'Fine\n');
    else
        ME.rethrow();
    end
end

end

function sub_test
sub_sub_test();
end

function sub_sub_test()
if rand>0.5
    error('test:error', 'Noooo!');
else
    error('test:notsobad', 'That''OK');
end
end

Example usage:

>> test_excc()
Error using test_excc>sub_sub_test (line 21)
Noooo!

Error in test_excc>sub_test (line 16)
sub_sub_test();

Error in test_excc (line 4)
    sub_test()

9           ME.rethrow();
K>> dbstack
> In test_excc at 9

Although the Matlab desktop environment prints the entire stack trace back to sub_sub_test, the debugger does not give me the possibility to go up the stack trace and debug inside this function.

I am aware of dbstop if caught error. However, this will debug into any caught error, which may be many if software makes heavy use of exceptions. I only want to stop on uncaught errors, but I want to stop where the error is generated — not where it's rethrown.

My question:

Upvotes: 4

Views: 656

Answers (1)

Richante
Richante

Reputation: 4388

I would guess that you cannot do this. As soon as execution enters the catch statement, dbstack will have to refer to that location inside the catch, so the information necessary to debug at the cause of the error is lost. ME.stack will tell you where the exception came from, but that isn't sufficient to debug at that location.

So I doubt you can solve the problem by doing something clever inside the catch. Looking at the documentation for catch, there also doesn't seem to be a way to do a java-style catch (ExceptionType ME).

There might be some hacky ways to solve this by using debug commands programmatically. For example, S = dbstatus saves the debug state and if there was a way to resume from a saved state, then you could attach this to the exception. But I can't find any documented way to do that.

Upvotes: 3

Related Questions