Better error reporting for stack overflow in type inference
: Line number of the function being inferred / call site
#49237
Labels
We've seen many errors in prod of the "stack overflow in type inference" variety, and one of the issues is that it's currently not actionable enough.
For example, we recently saw this error:
I truncated the message, but that
Tuple{}
goes on for a very long time.However, we currently can't figure out where in our code that is coming from!
It would be helpful to also report two locations, I think:
In this case, the function being inferred (2.) would be in
Base
somewhere, so not all that helpful. But maybe sometimes it would be.But (1.) would be extremely useful, so that we could figure out where this was coming from!
In fact, even better would be to actually dump the stack trace... if that was possible it would be a huge improvement. Thanks!
The exception is reported here:
julia/src/gf.c
Lines 324 to 327 in f32a5c6
Other examples of this exception reported in julia, here:
Internal error: stack overflow in type inference
when displaying an array with 511 dimensions #48813The text was updated successfully, but these errors were encountered: