Stream: t-compiler/wg-prioritization/alerts

Topic: I-prioritize #67375 ICE on unused generic in struct


triagebot (Sep 20 2020 at 23:48, on Zulip):

@WG-prioritization/alerts issue #67375 has been requested for prioritization.

Procedure

DPC (Sep 20 2020 at 23:49, on Zulip):

(this was marked as p-high before this wg was created, so re-prioritising to check if maybe need to reconsider the priority)

Hameer Abbasi (Sep 20 2020 at 23:56, on Zulip):

An illegal instruction sounds pretty high on the list for me, especially on stable. Why not critical?

DPC (Sep 21 2020 at 00:00, on Zulip):

for especially the newcomers: this was marked as p-high before this group existed, and back then p-high was the highest priority you could give. IIRC we did migrate some of the those accordingly, but either way p-high back then could equate to "modern day" :stuck_out_tongue: p-critical

lcnr (Sep 21 2020 at 05:36, on Zulip):

It looks like it doesn't emit "SIG_ILL" anymore

lcnr (Sep 21 2020 at 05:37, on Zulip):

the rest of this issue is the "classic const generics bug", where we get an ICE when referring to generic parameters inside of anonymous constants

lcnr (Sep 21 2020 at 05:40, on Zulip):

for those interested, the ICE happens because we do not supply the parent generics to anonymous constants, as doing so without feature(lazy_normalization_consts) causes cycle errors (we would have to compute a query to get the result of the exact same query)

This then means that when looking at the generic parameter T inside of inner the compiler gets confused as it did not expect any parameters here.

lcnr (Sep 21 2020 at 05:43, on Zulip):

Without the illegal instruction, this seems P-medium to me, as it is the exact same bug as #67945

lcnr (Sep 21 2020 at 05:44, on Zulip):

(also see #43408 for more discussion about this issue)

Joshua Nelson (Sep 21 2020 at 12:43, on Zulip):

Should it be closed as a duplicate then?

lcnr (Sep 21 2020 at 12:50, on Zulip):

yeah, tbh I don't see the point of keeping this open, it might be better to instead change this to E-needs-test with min_const_generics to see that we correctly handle it there

lcnr (Sep 21 2020 at 12:50, on Zulip):

as closures are fairly untested there iirc

triagebot (Sep 21 2020 at 13:06, on Zulip):

Issue #67375's prioritization request has been removed.

Last update: Apr 10 2021 at 21:30UTC