No error emitted for unknown identifiers about pythran

[serge-sans-paille]

Mmmmh, I assume this ends up with a clobbered c++ compilation error, am I right?
There is a implicit assumption in pythran that states that the compiler will not try to verify much things. If you're using polymorphic variables, it will try to generate polymorphic code, and will like fail.

I assume this can be implemented quiete easily, but it will not be a priority.

[mgaunard]

Errors emitted from the front end are better since they can reference the right line in the original Python code

[serge-sans-paille]

issuing #line is another alternative. this used to be implemented in pythran... anyway c++ errors are so difficult to grab...
Maybe I should restore it ?

[mgaunard]

I think it is good practice to emit an error as early as you can in the process, assuming finding out that error early doesn't require too much work

[serge-sans-paille]

I completely agree. Commit 5ea0d09 fixes this. Poor's man method: rely on pyflakes (if available) to perform a few checks.