Stream: t-compiler/wg-polonius

Topic: reboot meeting


nikomatsakis (Aug 27 2019 at 19:14, on Zulip):

Hey @WG-polonius -- I'd like to have a kind of "reboot" meeting where we discuss where we are, plan next steps, figure out who's around and for what, etc. Should we maybe use next week's slot for that? (This one being half over.) Or maybe we can just discuss in this thread a bit async?

For my part:

I'm still interested in keeping polonius on a "light boil" -- I probably won't have a ton of time to devote, but I'm happy we're still making progress. I can and will do the reviews of the initialization PRs, for example.

I still think the overall plan we've been following makes sense:

Albin Stjerna (Aug 27 2019 at 19:32, on Zulip):

Sounds good! I'm still interested in sticking around, but I also have no idea what will happen after I start my PhD. I will try to finish up the last bit of the initialisation (that is, reporting move errors), which shouldn't be a lot more work. I also would like to do the fact generation refactoring I described in the separate topic, if you (the collective you) feel it's a good idea.

nikomatsakis (Aug 27 2019 at 19:48, on Zulip):

To be honest I had some trouble picturing it

nikomatsakis (Aug 27 2019 at 19:48, on Zulip):

but I agree the current code is kind of spagetti

nikomatsakis (Aug 27 2019 at 19:48, on Zulip):

and I suspect your proposal will be an improvement

lqd (Aug 27 2019 at 22:06, on Zulip):

:wave: I'm slowly coming back online myself (and wasn't aware the meetings were back on already, sorry). I have been working on a summary of the current state of affairs, I'll try to finish it soon

lqd (Aug 27 2019 at 22:15, on Zulip):

(I'll mention it in more detail in the summary but I feel which of the 3 ways we choose to do the illegal subset relation errors has a big impact on the optimisation work; and also I've encountered some OOMs in the current fact generation and I wonder how to fix those and whether it has consequences on the approach or not)

Paul Faria (Aug 28 2019 at 04:31, on Zulip):

I'll be more available during late evenings. If I get enough of a heads up I can try to block time off from work to attend the next meeting. Is it scheduled at a regular time or on an ad hoc basis?

lqd (Aug 28 2019 at 10:56, on Zulip):

it used to be scheduled and in the compiler team calendar (but maybe not anymore I'm not sure)

lqd (Aug 28 2019 at 11:02, on Zulip):

@nikomatsakis here are the WIP notes I mentioned yesterday, let me know if you need more detail ?

Albin Stjerna (Aug 28 2019 at 12:29, on Zulip):

Also, concerning more videos etc; I will be required to hold a 30-minute presentation as part of my thesis work. I might try to make a decent recording of it if it would be of any interest to use as an introduction.

nikomatsakis (Aug 28 2019 at 12:34, on Zulip):

(I'll mention it in more detail in the summary but I feel which of the 3 ways we choose to do the illegal subset relation errors has a big impact on the optimisation work; and also I've encountered some OOMs in the current fact generation and I wonder how to fix those and whether it has consequences on the approach or not)

ah yeah I was thinking that (a) I had forgotten about the changes that @Aaron Weiss and I were discussing, which we never fully bottomed out and (b) the error work is important

Albin Stjerna (Aug 28 2019 at 17:04, on Zulip):

@nikomatsakisI made a chart, if that helps. Modules are in bold and facts in non-bold. Red and green is removed and added, respectively. Basically, my idea is to do everything we can do early in the process, and only put the things that absolutely need to happen inside of type-checking there (anything mapping variables or borrows to provenance variables). Apparently, outlives seems to take the output of type_check::type_check, so it can be pulled into the NLL module as well, just after the call to type_check().

Things that are grouped should take more or less the exact same inputs.

lqd (Aug 30 2019 at 15:20, on Zulip):

also, what does "reboot" mean in this context ? just re-starting the work after the holidays ?

lqd (Aug 30 2019 at 15:24, on Zulip):

(and if anyone has feedback on the notes I posted earlier, or needs more details or clarifications ? I'll be glad to update them before the meeting)

Albin Stjerna (Aug 30 2019 at 16:06, on Zulip):

@lqd I added a few comments, but other than that I don’t have anything in particular

lqd (Aug 30 2019 at 16:08, on Zulip):

I did see them yeah, thanks for those

nikomatsakis (Aug 30 2019 at 16:31, on Zulip):

also, what does "reboot" mean in this context ? just re-starting the work after the holidays ?

confirm

nikomatsakis (Aug 30 2019 at 16:32, on Zulip):

perhaps not the best term, I guess it has other connotations

Last update: Nov 15 2019 at 20:10UTC