@everyone hello all! meeting? Sorry I'm running a few minutes late.
let's all update the triage document
@lqd is it still true that html5ever runs out of memory? we should file an issue on that
yes it is, I don't think anyone worked on this yet
maybe @simulacrum has already opened an issue ?
There was something, let me look
there's https://github.com/rust-lang/rust/issues/51372 but that's unrelated
(also might be outdated)
true yeah, and I think @Matthew Jasper was looking at that
nothing else to my knowledge
(that I've filed)
ok, can someone file one?
anyway, you can see in the document, but first off: we did great last week
steady progress on performance
looking over the issues it seemed like it was time to focus some on diagnostics + ICEs
I triaged all the remaining issues: I put the "minor suggestions" into the Release Candidate box for now
I'll file something after checking
and focused on major areas of diagnostic fail
there are two: region error reporting and "unused mut", I think -- but @Matthew Jasper already has a PR around the "unused mut" lint errors :tada:
I opened a PR about the region ones that tries to move us closer to the objective, and I hope to (later today) filer various follow-up issues related to that
I think that we will soon have the results from @simulacrum's crater run
which will probably give us new stuff to triage and ICEs galore
so I expect we will have our hands full
oh yeah and some number of years ago in 1776 the US of A threw off the yoke of monarchic oppression
Crater claims 11 hours :)
but I plan to be working tomorrow anyway
that said, next week I am going to be on my way to Greece
so I probably wont' be around on Tuesday
I'll be there (in Greece) until August, but I will be working from there 4 days a week, so you probably won't notice the difference ;)
hopefully @pnkfelix will be back around next week though
(zulip is seriously lacking in the flags and feta cheese emoji department)
I blame the Unicode committee
What timezone is that?
PS I love these sorts of headlines: http://www.latimes.com/business/technology/la-fi-tn-emoji-q-and-a-20160125-htmlstory.html
gmt+2 I think
"Meet the shadowy overlords who approve emojis "
yes, it is GMT+2
Unfortunately, our website is currently unavailable in most European countries. We are engaged on the issue and committed to looking at options that support our full range of digital offerings to the EU market. We continue to identify technical compliance solutions that will provide all readers with our award-winning journalism.
so you may notice that difference =)
that said, I tend to work relatively late hours
when I'm in Greece
so as to overlap better with US, and to permit a trip to the beach in the AM
anyway, enough about that. I've been slow but I'll prepare a more detailed schedule in any case (just for my own sanity)
re: the ICEs and things...
I tried to put at least some mentoring instructions on most of the things I nominated, I think?
https://github.com/rust-lang/rust/issues/52028 for html5ever filed
with varying degrees of comprehensibility
I plan to write an internals post and tweet out the link, encouraging people to sign up if they have time -- but please if you want to grab an issue don't hold back :)
I would particularly like to see somebody start in on https://github.com/rust-lang/rust/issues/46908
@nikomatsakis what can I work on?
ah, perfect timing ;)
should I tackle this https://github.com/rust-lang/rust/issues/46908 ?
one of the issues — https://github.com/rust-lang/rust/issues/50467 — is mostly about reducing to a standalone test case. I thought @Jake Goulding you might enjoy that ;)
(I'm building a fix for 52027 locally)
@Santiago Pastorino please do
take a look at the instructions and see if they make sense (at least the high-level idea)
Is there more work you'd like me doing related to the region errors - you mentioned your PR today would open up a handful of opportunities; or would you prefer I just take on whatever?
there is definitely more work but I haven't opened up the issues yet
probably makes sense to do something else to start
as I guess it'd be easiset if that PR were to land first
but we can discuss in the meantime the follow-up plans?
also don't feel you have to focus on the E-mentor ones ;) if something else catches your fancy
@David Wood I forget, did you work on the 2-phase activations stuff? Oh, that was @Reed Koser I think
anyway this might be a small-ish but fun refactoring in that area
I don't think I did.
https://github.com/rust-lang/rust/issues/51351 is a bit hairier but at least tangentially related to regions
my mentoring instructions there were maybe a bit "notes to myself" ;)
I'll have a look at them and see what I come up with.
sounds good, well, that's all I got to say :)
not sure if anybody else has any notes or questions
is there more work related to liveness maybe ? (maybe the 40-50% left to grab on tuple-stress). I guess we'll also have the crater triage to handle soon
ah good point
so, I think there will be follow-up work on tuple-stress
one thing I was thinking about filing as an issue was experimenting with using sparse bitsets
or html5ever, which is basically a simili-ICE
instead of the dense bitsets we are using now
this would be a fairly minor tweak to the code
though it would invalidate @Santiago Pastorino's https://github.com/rust-lang/rust/pull/52007 PR (but if it pays off for tuple-stress, that seems ok)
the advantage would be that iteration is O(number of live things)
well plus if there are way more locals than live locals
then of course it's a win
a more radical thing would be to try and use :frog: somehow :)
I'd prob try the sparse bitsets first
just because it's easy and not disruptive :)
we mentioned before doing liveness in Polonius issues I think (at least, parts of it)
it was mostly brainstorming of course :)
yeah I mean eventually I'd prob like to move more and more of the dataflow-like analyses into polonius
it could also be an interesting issue (sparse bitsets) for other contributors eg in internals thread, twitter etc
(this is what I remembered from the Polonius issues linked to liveness, "emit def/use facts from rustc" + :frog: https://github.com/rust-lang-nursery/polonius/pull/45#discussion_r190516457)
if it's the end of the meeting, good evening / afternoon everybody :wave: