There's no general solution for the problem of repeatedly printing the same error

[sethfowler]

For a variety of reasons (fixed-point algorithms, compiler transformations that duplicate portions of the IR tree, utility code being invoke from multiple passes) we can end up printing the same error message many times. Loop unrolling is a place where this is biting us in a particularly nasty way.

There are ad hoc solutions for this problem at various places in the code; one example is the usage of iterCounter here:

https://github.com/p4lang/p4c/blob/master/frontends/p4-14/typecheck.cpp#L291

It would be good to have a more general solution to this problem. We should think about what the requirements are.

[mihaibudiu]

The simplest solution is to bail out from compiling. This works for errors, not for warnings.
A PassManager will stop after the first pass that signals an error by default. So will a RepeatedPass.
So the only issue is within a pass, which should be tractable.

(I am not aware of any loop unrolling code that is being run in the front-end.)

[sethfowler]

Bailing out would work for errors, that's true. It's nice to report as many errors as possible, though, to reduce the number of edit-compile cycles a P4 developer has to go through.

[mihaibudiu]

I will close this issue because it is too general. An example would be more useful.