Hi everyone! I am probably not going to be able to attend the meeting, but I figured I could write a quick report anyway. I'm mainly working on the presentation for my thesis and [non-Polonius stuff]. Once The Great Renaming and my Other Things have settled down, I figure I'll go into fixing Niko's concerns about initialisation to make the tracking of move paths precise. For that, I would need some way to do verification, for example the Rust unit tests with Polonius (that I don't know how to run).
After that comes move errors, which I have code for in pencil on a sheet of paper somewhere. To unblock that would also require some way of verifying it (passing the results on to rustc and running the compiler's unit tests?), and a decision on how to report errors that are not invalidation x live loans-errors (that is, the current set of errors). Of course, I can just make something up and fix it retroactively.
Other things I would like to work on in the reasonably near future is:
--benchmark-factsthat would change Polonius' behaviour to take a path containing nll-facts for a number of crates, and produce results like
crate_folder_name,fn_name,initialisation_runtime,liveness_runtime,loan_vilations_runtime, or something like that.
those are all of great interest and very useful :thumbs_up:
(also I like that The Great Renaming™ terminology is catching on ;)
@nikomatsakis I see you're firefighting some async issues, do you need to skip today's meeting ?
hello friends from @WG-polonius :wave: — Albin mentioned they might not attend so it's likely going to be fairly sparse in here :)
I'll give my update about this week so at least people can ready about it async
I was supposed to read Albin's thesis and give feedback but I didn't manage to do so
hey niko :)
Yeah, I don't have a ton to report -- I talked to @lqd some last week and brought myself up to speed on what they'd done on equality -- I hope to put in a bit more time this week to look at subset errors
oh I forgot about my PR proposing names
I have some unread messages in this stream :)
one or two at most :)
so I'll continue, I have I think a good enough overview of inputs to be reviewed
as for naming I have started renaming regions to origins
(I'll add all the links later)
I've updated the current status of the rustc test suite, and opened a PR on the trivial cases
doing so I found there was a new OOM, which I also investigated
and it's very similar to the previous one, and also what we experienced with NLLs
e.g. here a lot of arguments/temporary variables coming from
assert_eq! and the likes
and that's pretty much it :)
I plan to continue on The Great Renaming this week, adopting the new nomenclature in the "polonius inputs" doc and opening a PR for that
and read Albin's thesis
thanks @lqd :heart:
Sounds like we should land https://github.com/rust-lang/polonius/pull/124 ?
(I've added links the previous messages in case)
yeah let's do it
I'll get used to Nodes vs Points ;)
yeah, hmm, we'll see if I can
that is probably the least important one, I'd be fine to revert it, as long as we avoid single-letter names
which I think we sort of agreed to do anyway
about thos single letter names
I see @lqd you opened polonius#125
was it about them in generic types ?
or variables as well
mm I don't know, I was thinking also in the book
e.g., in "datalog rules"
I guess everywhere :)
was it about them in generic types ?
I'm always unsure about this -- single letters are sometimes kind of nice to clearly separate out "generics", but I can't see a good argument for it
so for L0 I should eg switch to Loan0 everywhere ? comments, datafrog closures, etc ?
yeah, like that
Loan0 seems sort of strictly better, no?
alright I'll take care of this this week then
Albin also posted an interesting message at the top of this thread about possible tasks we'd need to do
but in general we can talk more about this async for sure
these and the other ongoing topics
looks like we're close to the end, if so thank you for your time everyone :) have a nice evening/afternoon :wave:
(I'll still be around)