Stable compilers don't emit ICE files?

[saethlin]

I don't know for sure if this a bug or an unexpected intentional decision.

It looks like, based on some experimentation and recent ICE reports on stable such as https://github.com/rust-lang/rust/issues/132240, that the stable compiler does not write and mention ICE files. I don't want that behavior, because some fraction of the time people try to be helpful and abbreviate the compiler's output, and instead omit all the important information.

[jieyouxu]

ICE disk dump MCP: https://github.com/rust-lang/compiler-team/issues/578 Previous compiler discussion: https://rust-lang.zulipchat.com/#narrow/stream/233931-xxx/topic/Store.20ICE.20backtraces.20to.20disk.20and.20point.20en.E2.80.A6.20compiler-team.23578 Original PR: https://github.com/rust-lang/rust/pull/108714

Haven't figured out the rationale for why the ICE dumps are nightly or dev only. cc @estebank maybe you know?

[estebank]

It was explicitly made nightly-only so that it would have time to "bake" and minimize the likelihood of this infrastructure breaking stable users due to some hypothetical obscure bug that only affected ICE dumping. I would be in favor of enabling it in stable now, for the same reasons @saethlin wants it enabled. I was initially thinking that we should wait until the cargo gc work was finished before moving this to stable, but at the same time I feel it should be fine to change the path we write to after the fact.

CC @yaahc, as she's working on this space and might have thoughts. I know that we were looking at tweaking the setting and override for these.

[compiler-errors]

Why is this pinned

edit: I'm gonna unpin it. It's not particularly user-facing or anything.

[yaahc]

I have no objections to us enabling the ice dumps on stable. It still wouldn't be considered a part of our stable interface so even if we do run into problems down the line with disk usage (which I don't find likely) we can always disable it again while we get that sorted.