Plugin - Polysemy

Hi there, is it still worthwhile to use polysemy-plugin?

Why it shouldn't be anymore?

maybe because of the panics with 8.10

Ah, I see - I'll try to take a look at it

you should take a look at my pr! I probably messed something up

https://github.com/polysemy-research/polysemy/pull/337

That shouldn't be it (like please, please GHC, you wouldn't do this to us :sweat_smile: )

but it's already merged

yup, and there are problems with 8.10 :sweat_smile:

that pr was so that it actually compiled in the first place

I mean, it shouldn't be the reason

ohh

You know what's interesting? Part of plugin concerned with types works in typechecker solely, but error is related to information created during core transformation that removes unused function arguments...

We need to try it without additional core passes added by the other part of plugin

(They come from GHC, so maybe it's actually their problem? :big_smile: )

btw I have a project (private, so I can't share, large, so I can't minimise) in which I often get "stuck type family bogus proof etcetc" errors

:(

from the plugin

e.g. I just made a let binding in a do block, of which the body was just a single hole, and I got the same thing

same

same also regarding the "large so I can't minimise"? :sweat_smile:

but like 99% in ghcide but not ghci

well.. my errors are also in hls, I haven't actually tried it in ghci..

:thinking:

Yeah, that's a separate problem - I think that error comes from plugin (have to check), so maybe we do some incorrect assumptions about the input

no I thought that was just unavoidable. I think there is a github issue on the topic

in polysemy or in ghcide?

polysemy

https://github.com/polysemy-research/polysemy/issues/363

for me it often happens if I have a wrong number of arguments for an effect ctor

I guess this is a pretty small example

there ya go

but it's also some interaction with how ghcide runs the compiler I guess..

hm yeah

Okay, so I've disabled core passes... and problem solved :joy:
@Sandy Maguire seems like we missed it because they were written to only run on 8.10

lol awesome!!! :rocket:

(Problem solved = zoo compiles - we may need to try it in other projects)

But it seems like it's really some problem in GHC after all?

@Georgi Lyubenov // googleson78 @Love Waern (King of the Homeless) does it make sense to remove these? We weren't using them until now, and as we already know we can't get past module boundaries just by making GHC optimize more heavily

who pings me???

Just your past GHC plugins haunting you :stuck_out_tongue_wink:

To add some context, disabling extraPhases in https://github.com/polysemy-research/polysemy/blob/6b660eaadca54a34aba2963707175233e5a7da7f/polysemy-plugin/src/Polysemy/Plugin.hs#L97 solves https://github.com/polysemy-research/polysemy-zoo/issues/65

I think it's fair to remove/disable them until someone has the time to invest in "going faster"

especially since it's not actually a regression

https://github.com/polysemy-research/polysemy/pull/382