Internal Fulfillment Context assertion tripped, needs more investigation

[jroesch]

I'm not sure if this is a bug or not, needs further investigation. It appears that by reusing the fulfillment_cx here we incur more obligations and later trip an assertion.

The two possibilities I see is:
- normalization is not actually fully happening and we have a bug else where
- we are adding a duplicate bound into the list causing its size to change.

[steveklabnik]

Did you ever investigate further?

[jroesch]

@steveklabnik I'm currently rewriting a large chunk of the fulfillment context that I believe will make this a non-issue, but we should probably keep this open for the time being if is okay with you?

[steveklabnik]

👍

[soltanmm]

@jroesch There's a FIXME comment that references this issue at present. Is that supposed to be there?

[jroesch]

@soltanmm this is references a problem with the current fulfillment context design, where we can't re-use fulfillment contexts because they have no transactional support, this was fixed on my branch refactoring the inference engine. I'm currently trying to rebase this with #30533, still need to touch base with Niko about it though. Was with family for my grandmother's 90th birthday that past couple days, home again and back to work.