Hello @T-compiler :wave:
I'm brewing up some coffee, so brb :)
#52966 -- New deny(const_err) in unreachable code
Hmm, you're right -- rustc 1.10 through 1.25 do lint this as a warning. 1.26 through 1.28 do not, and now 1.29 lints it again, denied by default.
so I guess the question is .. do we do anything here?
This is an expected regression after miri
I didn't implement the linting for casts, now I did
I think the only real 2 options are
do we leave as is, or do we make
const_err warn-by-default again
sounds like "leave as is" is probably fine
next up: #52849 -- A pretty big performance hit after a recent change
specifically, some kind of CTFE compile-time regression?
yea, @RalfJ fixed most of it
perf tests will appear in the next ~5 hours
still some weird edge cases, but we have benchmark canaries now
so we'll leave it as is for now
next: #52390 -- Segfault during tests on rust 1.27.x on macOS
@nagisa is this one of those things you were talking about in the sigsegvs topic?
feelings in last week's meeting were to close this week XD
oh,I see, this is the TLS thing
ok, let's close
the issues I’ve talked about were about the compiler itself crashing
#52141 -- [Rust 2018] pub use some_crate as some_name breaks glob imports
If someone can investigate this now, please do.
I'll be able to do it in 3-4 weeks.
seems like we might want to find someone to look
cc @eddyb — are you still working on the name resolution code as part of EP2 work?
although there is https://github.com/rust-lang/rust/pull/52923
it looks like @eddyb is working on this, let's assign to them
next is #52140 which looks very similar?
#52140 -- [Rust 2018] pub use can't export crates
I'll assign to @eddyb too :)
since they don't seem to be around to defend themselves :devil:
#52092 -- Recent 12% perf regression in unused-warnings
I have my machine back and can do some profile now
I've done nothing though
I guess I'll give that a go today
#51608 -- Code fails to link on macOS with incremental compilation
last week we wanted to determine whether this is really p-high
since none of the stakeholders responded, I guess it's not?
I guess we can downgrade until somebody complains
but -- is this a serious problem that might be affecting others?
or is it something very servo specific?
it seems to be a very specific case, with explicit linkage for a given global
since you get a linker error, I think people would have complained more if they ran into it often
ok, downgraded, shall I change you to P-medium?
er, I meant
shall I unassign you?
#51212 -- resolve suggestions do not use
crate:: when enabled
so, the title is not fully accurate anymore
I believe that there was a quick fix for using
but what won't work is getting suggestions from other crates when
extern crate is not present (right @qmx ?)
cc @simulacrum I'm not sure if this should be P-high...
working on the full solution atm, after your pointers
P-high because edition critical
We can downgrade to P-medium, but we want this for the edition for sure
ok, I'll leave as is, @qmx and I have been discussing some and there's been progress
maybe we should make it P-medium but put on a milestone
if it's not already
that works for me
oh, it is on one
ok, @qmx, let's sync up later today anyhow
there we go
Compiler crashed with cannot create local mono-item #50865
(man, this far into the meeting before I realized I can copy-and-paste exactly what GH displays...)
lots of backlog here, I remember this now
seems like somebody ought to take charge, but not clear who that person should be
well, it's RC milestone, let's levae it for now
rust 1.25.0 internal compiler error #49482
I think @mw that we don't necessarily have metadata etc for all things that get a
could that be the problem?
o yeah, I looked into this a little but it's weird
they should be in the DefPathTable
which is all that should be needed here
ok, well, I guess for now I will leave assigned to you?
I still think it has to do with proc-macro crates being weird about metadata
seems like a likely cause
yeah, unless you want to see progress :P
is it hard to reproduce?
there's GH repo that you just need to clone
I guess the question is whether we believe it P-high or not
it is a regression
I'll try to look into it some more next week
hard to tell how widespread
let's leave as is for now
syntax_pos::symbol::Symbol::gensym() is incompatible with stable hashing. #49300
what should we do about that :point_up: ?
@pnkfelix writes that maybe we should downgrade to P-medium
I believe this was one of those "I don't want to forget about it" sort of things
I'm kind of inclined to agree we should downgrade
yes, I'm OK with that
there's an assertion in the code now that should catch this early, I think
Tracking issue for RFC 2093: Infer
T: 'x outlives requirements on structs #44493
FCP is complete
I guess we should get someone to stabilize
I bet that @toidiu would be up for that :)
we should port the compiler to use it too .. oh, we have to wait until beta
/me can't wait to stop typing
'cx, 'gcx: 'tcx, 'tcx: 'cx
we should probably split off a separate issue for the new feature gate?
ok, that's all of P-high
Wildcard imports do not seem to propagate as expected for tuple-like struct types having at least one non-pub field #53140
kind of looks like... T-compiler?
bisected to https://github.com/rust-lang/rust/pull/52555
I guess I'll mark as P-high, @Vadim Petrochenkov already assigned themselves...
everything else looks properly categorized
Regression importing log’s warn! macro inside of include!()ed file #53205
I'm not really clear on what this is but @Vadim Petrochenkov seems to be on it
P-high I guess, as it's a regression
oh dear, many
Internal compiler error: unexpected panic: thread 'main' panicked at 'no entry found for key' #53146
but @Esteban Küber reports "Happens on stable 1.28. Doesn't occur on beta, nightly or 1.27.2."
some kind of impl trait bug I guess
I'm inclined..to close?
/me getting lazy I guess
the question is whether we would do a point release for this
and I think not
Hm, probably no
probably good to bisect though
I'll assign to myself for bisection
ok, I've marked as P-medium in the meantime
rustc doesn't output anything but exits with 101 and doesn't emit any files #52913
regression introduced by @Sunjay Varma in https://github.com/rust-lang/rust/pull/45904, according to @Esteban Küber
or perhaps by me, since I was reviewer :P
maybe I misunderstood the comment
I guess queston is how to prioritize this...
P-high I suppose. seems bad. Even though GATs are "not really working", someone could stumble on this by accident.
shouldn't any use of a GAT be considered an error until they're actually working?
or are there some cases where they work enough that people might be using them?
they don't work
the bug is that rustc emits nothing at all
vs some kind of error
I'm not really clear on why that is :)
rustc fails to raise rlimits for stacksize on mac for doctest runs through make #52801
I have no real idea what to do about this...
is it a regresson really?
I guess it occurs because of the stack guard?
oh, hmm, is the problem that we are trying to "raise" the stack limit but it's already been raised even higher...?
anyway i'm inclined to call this P-medium
ok, that's the list of new regressions
looking at I-nominated issues
one thing is a PSA: we're starting to land PRs that add
#![feature(nll)] to all the rust crates
tracking that effort in #53172
Rustdoc ignores traits implemented inside of a function #52545 is nominated
seems like fix is underway
the other nominations all seem stale
I have a thing nominated by the tools team (I think) for the compiler team. The clippy 1.0 RFC is currently waiting on us to decide on lint uplift of correctness lints: https://github.com/rust-lang/rfcs/blob/b9c8471887f308223c226642cad3a8290731b942/text/0000-clippy-uno.md#correctness-deny
Should I create an issue with all the lints and put it into FCP?
i.e., to go through that list and decide which lints we will uplift?
that sounds good
perhaps T-compiler + T-lang
ok, will do
or at least just cc the lang folks
not really clear where that stuff falls to me
doens't matter too much
yes please :)
Fix NLL migration mode so that reports region errors when necessary. #53045
that one is not approved
but seems like it clearly should be
ok, we're done!
anybody want to raise anything else?
thanks all <3
@nikomatsakis The regression you mentioned where no output is produced was already known to us from the very beginning: https://github.com/rust-lang/rust/issues/47206
We thought it was weird and knew we should look into it but then no one ever did
@Sunjay Varma oh :) forgot about that!
where in assoc ty impls should probably be behind the feature gate too, right?
so the example in the new issue shouldn't even get to that stage
might be an oversight, not sure