regression: binop(?) type inference failures

[Mark-Simulacrum]

A bunch of errors around type inference failure detected in the crater run -- I would like us to track down the PR causing these since I'm not finding an obvious reason off hand:

• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/mwc10.dt-backup/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/DaTa-.advent-of-code-2020/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/MikkelPaulson.advent-of-code-2019/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/bfollek.advent2018rust/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/daniellockyer.wrp/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/proelbtn.chacha20poly1305/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/qinxiaoguang.rs-lc/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/ruudjelinssen.aoc2020/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/reg/roots-0.0.7/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/reg/segsource-0.2.0/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/desbma.motd/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/l0s.advent-of-code-rust/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/reg/dent-0.4.1/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/GCYYfun.OSTEP_RUST/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/gh/botwob.AdventOfCode2019/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.58-1/beta-2021-11-30/reg/zero_pass_backend-0.0.7/log.txt

It seems plausible this is the breakage noted in #90904, but we should confirm.

[Mark-Simulacrum]

We will want to do a re-run of beta crater once #91484 or equivalent is beta-backported to confirm the majority/all of these are fixed before closing this, but generally it's likely that they will be.

[BGR360]

I just did some bisection the other day; it's pretty painless on my M1 pro macbook (~2-4 minutes per build). If this still needs bisection, and somebody can give me a range of commits, I'd be happy to try to bisect it.

[workingjubilee]

Since each regression is likely due to #89167,
and each is likely fixed after #91484
we want to check, for each regression:

• does this build on e90c5fb?
• does this build after 032dfe4 but before 11fb21f?
• does this build after 11fb21f?

And any curiosities can be handled from there.

[BGR360]

For the regressions starting with reg/, where do I get those? The other ones have git repos I can clone (gh/), but I don't intuit what "reg" stands for.

EDIT: oh, does it just mean reg-ular crates.io crates?

[BGR360]

Here's what I got with the GitHub ones.

e90c5fb:

==== SUCCESSFULLY BUILT: ====
DaTa-.advent-of-code-2020
GCYYfun.OSTEP_RUST
MikkelPaulson.advent-of-code-2019
bfollek.advent2018rust
botwob.AdventOfCode2019
daniellockyer.wrp
l0s.advent-of-code-rust
mwc10.dt-backup
proelbtn.chacha20poly1305
qinxiaoguang.rs-lc
ruudjelinssen.aoc2020
==== FAILURES: ====
desbma.motd

3e018ce (one merge after 032dfe4):

==== SUCCESSFULLY BUILT: ====
==== FAILURES: ====
DaTa-.advent-of-code-2020
GCYYfun.OSTEP_RUST
MikkelPaulson.advent-of-code-2019
bfollek.advent2018rust
botwob.AdventOfCode2019
daniellockyer.wrp
desbma.motd
l0s.advent-of-code-rust
mwc10.dt-backup
proelbtn.chacha20poly1305
qinxiaoguang.rs-lc
ruudjelinssen.aoc2020

11fb21f

==== SUCCESSFULLY BUILT: ====
DaTa-.advent-of-code-2020
GCYYfun.OSTEP_RUST
MikkelPaulson.advent-of-code-2019
bfollek.advent2018rust
botwob.AdventOfCode2019
daniellockyer.wrp
l0s.advent-of-code-rust
mwc10.dt-backup
proelbtn.chacha20poly1305
qinxiaoguang.rs-lc
ruudjelinssen.aoc2020
==== FAILURES: ====
desbma.motd

Not sure what's up with desbma.motd. I made sure to check out the same revision of the repo that was used in the crater logs, but it failed to compile even on stable (with a different error than that in the crater logs). Maybe the author force-pushed that commit or something.

[workingjubilee]

Thank you, it looks like it was in fact that PR then and that did in fact fix it, it seems.
Ah, I suppose we could also check 477fd70, but I will likely do that when I try to reimplement splatting.

[emilyalbini]

Hmm, what's the status of this?