Stream: t-compiler/wg-polonius

Topic: meeting 2020.03.03


lqd (Mar 03 2020 at 19:54, on Zulip):

an early :wave:

lqd (Mar 03 2020 at 19:59, on Zulip):

as mentioned in last week's topic, I updated the ui test analysis document I said I'd do

lqd (Mar 03 2020 at 20:01, on Zulip):

(I also managed to find the braindump/roadmapesque document I said I'd find, but didn't yet manage to update it though)

lqd (Mar 03 2020 at 20:01, on Zulip):

the update to polonius 0.12 landed earlier #69482

lqd (Mar 03 2020 at 20:03, on Zulip):

the bit of bad news is I think move errors are buggy in polonius, at some point I was wondering whether it was related to fact generation in this PR, but deactivating what I think were the different facts not related to renaming didn't seem to change much

lqd (Mar 03 2020 at 20:04, on Zulip):

(that's why I also didn't close the PR per se, I haven't been able to investigate but I think it's in polonius)

Albin Stjerna (Mar 03 2020 at 20:05, on Zulip):

That would make sense to me

lqd (Mar 03 2020 at 20:05, on Zulip):

the 3 "maybe-initialized" tests are part of it, but I've also seen move errors in our polonius inputs which shouldn't have any

Albin Stjerna (Mar 03 2020 at 20:05, on Zulip):

That there's something wrong with the rules

Albin Stjerna (Mar 03 2020 at 20:05, on Zulip):

Ok, that's concerning

Albin Stjerna (Mar 03 2020 at 20:05, on Zulip):

Clearly I didn't test well enough

lqd (Mar 03 2020 at 20:05, on Zulip):

(that could be related to facts but I don't think so)

lqd (Mar 03 2020 at 20:06, on Zulip):

even though I'm not sure at all and haven't been able to investigate so it's probably not wise to speculate :)

lqd (Mar 03 2020 at 20:07, on Zulip):

(despite my limited ability to work outside lunch break at work, I had been working on another thing which I was excited about but trying placeholder loans destroyed it as well — all in all I don't like placeholder loans as much as when I thought we wouldn't need to materialize them at all points in the cfg :)

nikomatsakis (Mar 03 2020 at 20:07, on Zulip):

wave

lqd (Mar 03 2020 at 20:08, on Zulip):

the move errors false positives are probably easier to find than the maybe-initialized lack of errors, since we kinda only have why-provenance and not why-not provenance :)

lqd (Mar 03 2020 at 20:10, on Zulip):

and of course we'd still need to make rustc emit them, which would be the easiest way to test move errors, really

nikomatsakis (Mar 03 2020 at 20:10, on Zulip):

I'm skimming this ui test analysis document

lqd (Mar 03 2020 at 20:10, on Zulip):

start at #34

lqd (Mar 03 2020 at 20:10, on Zulip):

the rest are older

nikomatsakis (Mar 03 2020 at 20:11, on Zulip):

woah :)

nikomatsakis (Mar 03 2020 at 20:11, on Zulip):

ok

lqd (Mar 03 2020 at 20:11, on Zulip):

(I should have updated it earlier sorry, just subset errors changed things a bit, and involved more closures, I've tried to add most new ones)

nikomatsakis (Mar 03 2020 at 20:12, on Zulip):

lqd said:

the 3 "maybe-initialized" tests are part of it, but I've also seen move errors in our polonius inputs which shouldn't have any

so 36-38

nikomatsakis (Mar 03 2020 at 20:12, on Zulip):

is maybe where I should start?

lqd (Mar 03 2020 at 20:12, on Zulip):

I think those are move errors bugs

lqd (Mar 03 2020 at 20:12, on Zulip):

you probably shouldn't bother with this document with your limited time :)

nikomatsakis (Mar 03 2020 at 20:13, on Zulip):

maybe so :) but it seems like the bug here is mostly 38, right?

lqd (Mar 03 2020 at 20:13, on Zulip):

although #41 and others are "interesting"

lqd (Mar 03 2020 at 20:13, on Zulip):

I don't think so no

lqd (Mar 03 2020 at 20:13, on Zulip):

I think the lack of error in 38 invalidates 36 and 37

lqd (Mar 03 2020 at 20:14, on Zulip):

if there's no error to fix, there's no "unexpected pass" is what I thought, but I'll need to investigate those :)

lqd (Mar 03 2020 at 20:14, on Zulip):

(and maybe the false positives elsewhere are an easier way to find issues in the rules, if there are any)

nikomatsakis (Mar 03 2020 at 20:16, on Zulip):

well

nikomatsakis (Mar 03 2020 at 20:16, on Zulip):

I mean what you're saying is

nikomatsakis (Mar 03 2020 at 20:16, on Zulip):

fixing 38 may mean we start giving errors for 36/37

nikomatsakis (Mar 03 2020 at 20:16, on Zulip):

but none the less 38 is the one giving the output that we don't expect presently, right?

lqd (Mar 03 2020 at 20:16, on Zulip):

yeah

lqd (Mar 03 2020 at 20:17, on Zulip):

all 3 are unexpected

lqd (Mar 03 2020 at 20:17, on Zulip):

it could be something minor for sure

Albin Stjerna (Mar 03 2020 at 20:19, on Zulip):

For example an off-by-one in overlaps

Albin Stjerna (Mar 03 2020 at 20:19, on Zulip):

I.e. confusion about at/before semantics

Albin Stjerna (Mar 03 2020 at 20:20, on Zulip):

It's not as if I haven't messed that up before

lqd (Mar 03 2020 at 20:20, on Zulip):

so Niko do you want to unschedule this meeting soon and try to move to a more periodic work week ? I presume pushing move errors over the finish could be a topic :) (if we haven't managed to find the issue by then ofc)

lqd (Mar 03 2020 at 20:20, on Zulip):

@Albin Stjerna I'll try to look for it as well, these are pretty subtle sometimes (and I know very little about it)

lqd (Mar 03 2020 at 20:21, on Zulip):

(also comparing 0.11 and 0.12 could lead to clues, even after our renaming)

nikomatsakis (Mar 03 2020 at 20:23, on Zulip):

lqd said:

so Niko do you want to unschedule this meeting soon and try to move to a more periodic work week ? I presume pushing move errors over the finish could be a topic :) (if we haven't managed to find the issue by then ofc)

I think that's probably a good idea

nikomatsakis (Mar 03 2020 at 20:23, on Zulip):

the question is what frequency to use for periodic work weeks

lqd (Mar 03 2020 at 20:23, on Zulip):

I'll try to look for this this week, and will report asynchronously if I find anything Albin, I know you're pretty busy as well

nikomatsakis (Mar 03 2020 at 20:23, on Zulip):

(and definitely y'all should feel free to keep working even if I'm not paying attention)

lqd (Mar 03 2020 at 20:24, on Zulip):

sure

lqd (Mar 03 2020 at 20:24, on Zulip):

as to frequency, I'm not sure :)

lqd (Mar 03 2020 at 20:25, on Zulip):

most of this depends a lot on y'all schedules

lqd (Mar 03 2020 at 20:26, on Zulip):

(that is, I'd probably take some vacation days for that)

nikomatsakis (Mar 03 2020 at 20:27, on Zulip):

I think maybe we should look to April

lqd (Mar 03 2020 at 20:27, on Zulip):

April sounds good

lqd (Mar 03 2020 at 20:29, on Zulip):

(I might try to prototype moving away from placeholder loans, they're killing me :)

Albin Stjerna (Mar 03 2020 at 20:30, on Zulip):

April sounds good for me too, I think

lqd (Mar 03 2020 at 20:30, on Zulip):

in any case, no need to settle on a work week date tonight, tentatively around April, if that's alright with you 2 (I know you both have little time) and we'll see how things progress until then

lqd (Mar 03 2020 at 20:32, on Zulip):

btw if that's all we wanted to cover, we're at 30mins, if so, thanks a lot for your time, and have a good afternoon/evening :)

lqd (Mar 03 2020 at 20:33, on Zulip):

(I'll still be here)

Last update: Apr 03 2020 at 17:20UTC