Chore: bump to lts-20.0 #3642

mitchellwrosen · 2022-11-29T21:25:29Z

Overview

Redoing #3631, this time landing on GHC 9.2.4

aryairani · 2022-12-01T02:01:28Z

Superficial HLS test seems okay. I do have to open the package containing the definitions I want to jump to before I can jump to them, which seems kind of pointless, but I'm not sure it didn't already work that way.

aryairani · 2022-12-01T02:02:13Z

Reminder to create corresponding PR for enlil.

aryairani · 2022-12-01T12:22:15Z

stack.yaml

- lsp-1.5.0.0
- lsp-types-1.5.0.0
- text-rope-0.2@sha256:53b9b4cef0b278b9c591cd4ca76543acacf64c9d1bfbc06d0d9a88960446d9a7,2087
+  commit: 3b26f1e6c279e9bb5cd052c34c852aa8f8cfa17e


Sanity check: did you ensure color renders properly in a Windows terminal, or what commit is this?

I'm testing it, but it will take a while.

This commit is morally the same: the PR that we were pointing at was rebased by the author atop the latest main so it continues to build. The PR still isn't merged :/

This is what I get; let me check again on trunk.

Oh. Sorry for making you do all of that, but I understand now.

I too went looking for "what is commit d6c2643b0d5c19be7e440615c6f84d603d4bc648?" when trying to fix compiler errors with haskeline due to the bump, and I saw it in the history of the open PR we were depending on:

But I got the history wrong. Looks like the author was not just force-pushing over their PR to keep it atop the main branch, they changed the implementation, too.

Meanwhile @ChrisPenner (maybe independently) discovered that the original approach was correct, so pointed us at d6c2643b0d5c19be7e440615c6f84d603d4bc648.

So I believe the correct move is to grab haskeline master (since it builds), and revert haskell/haskeline@f827f10 on top of it (which is how d6c2643b0d5c19be7e440615c6f84d603d4bc648 came to be).

@aryairani I've updated this dependency. Now it's pointing at our unisonweb/haskeline fork, branch lts-20.0. The contents of the branch are just judah/haskeline HEAD, plus this commit, which is equivalent to the previous commit we were using (although now rebased on top of judah/haskeline HEAD, because the old commit didn't build against lts-20.0).

Some confusion: we have a unison branch (the default branch) in our unisonweb/haskeline fork, 7 commits ahead and 13 commits behind aryairiani/haskeline. I first tried merging judah/haskeline HEAD into this, but got a ton of conflicts. Is this branch just dead code at this point?

Is this branch just dead code at this point?

I would guess so.

screenshot of 20dacc9 courtesy of @tarikozkanli

aryairani · 2022-12-01T12:23:49Z

unison-cli/src/Unison/CommandLine/Welcome.hs

@@ -166,7 +166,7 @@ authorSuggestion :: P.Pretty P.ColorText
 authorSuggestion =
  P.newline
    <> P.lines
-      [ P.wrap "📜 🪶 You might want to set up your author information next.",
+      [ P.wrap "📜 You might want to set up your author information next.",


Can we not use the 🖊️ character anymore?

I haven't tested on GHC 9.2.4. On GHC 9.0.2, no, which was weird. But I didn't investigate too closely because I thought putting two emojis in a row here was a mistake, anyway.

Should I try putting it back?

I dunno, I guess it's supposed to be writing on paper, not sure if the paper alone makes sense. @rlmark ? :)

unison-hashing-v2/src/Unison/Hashing/V2/Term.hs

new regression discovered

ChrisPenner · 2022-12-02T16:25:48Z

Has anything changed on the points listed from the last time I attempted moving to 9.2?

#3304

Don't want to upgrade to (yet) 9.2.3 because :
Wingman isn't yet supported via HLS on 9.2*, which means no case-splitting functionality. However there's active work to split off that functionality to a new plugin which would be supported for 9.2.3: haskell/haskell-language-server#2971

It's okay if we decide that's not a priority, but we should consider it I think 😄, I use the case-splitting feature very often.

mitchellwrosen · 2022-12-13T19:18:21Z

It's okay if we decide that's not a priority, but we should consider it I think smile, I use the case-splitting feature very often.

I think it's mostly up to you then whether we upgrade now or later. Personally I've never gotten that feature to work with any hls.

aryairani · 2022-12-23T15:03:07Z

Current status is:

windows prompt is still broken (cc @mitchellwrosen)
waiting on either case-splitting Wingman support, or for us to declare we don't care enough to keep holding off (cc @ChrisPenner)

ChrisPenner · 2023-01-03T20:15:51Z

The windows prompt seems important,

as for wingman, I'll miss the pattern splitting, but it wouldn't kill be me to go without it, copilot can help fill that gap I suppose.

I tried out the GHC 9.2.4 build and I was hoping it would speed up GHC compile times on M1 more than it actually did, so that's a bummer, but still probably good to keep relatively up to date with GHC 👍🏼

ceedubs · 2023-01-04T13:07:43Z

For what it's worth, I'm hoping that these memory changes in GHC 9.2 will help address some memory issues in Unison Cloud related to the GHC runtime using way more memory than the allotted heap.

mitchellwrosen · 2023-04-18T15:11:05Z

This PR is old and dead, which is a shame. It's sometimes quite a lot of work to bump the LTS, and we are on quite an old GHC at this point, so it would be good to upgrade.

I do sometimes wonder if it's worth it for our small team to support Windows. If even basic libraries like haskeline just don't work out of the box, and it's on us to debug and figure out - well, maybe the Haskell ecosystem just isn't good enough for cross-platform stuff 😓

ChrisPenner · 2023-04-18T15:22:00Z

I agree that we should find a way to upgrade soon, the longer we wait the further behind we get. It's likely worth a week of off-sprint time to figure out the haskeline thing at this point.

aryairani · 2023-05-19T13:12:55Z

We do want to support Windows, but yeah agree that the situation is pretty ridiculous.

aryairani · 2023-05-19T18:36:18Z

@mitchellwrosen Ok, when I build this PR with

- github: judah/haskeline
  commit: d6c2643b0d5c19be7e440615c6f84d603d4bc648

(same as on trunk), then UCM works with allow-newer: true; without it, I'm just missing the bumped bounds that the newer versions of Haskeline have.

In the dependencies for haskeline-0.8.0.0(-terminfo):
  base-4.16.3.0 from Stack configuration does not match >=4.9 && <4.15 (latest matching version is 4.14.3.0)
  bytestring-0.11.3.1 from Stack configuration does not match >=0.9 && <0.11 (latest matching version is 0.10.12.1)
needed due to unison-cli-0.0.0 -> haskeline-0.8.0.0

I confirmed that rebasing our preferred Happy Windows commit onto the latest Haskeline, and building that with the new LTS is no longer happy on Windows, so I guess something else broke in Haskeline that breaks the proposed-and-then-retracted fix from @. this was what you had done too that also didn't work back in December (20dacc9)

I guess we should just use the commit we know, put it on a branch to satisfy #3945, and bump the bounds ourself?

aryairani · 2023-05-20T22:50:11Z

superseded by #4008

mitchellwrosen added 7 commits November 21, 2022 15:40

⅄ 22-11-21-remove-strings → 22-11-21-bump-lts

a499151

bump from lts-18.28 to lts-19.0

106789d

move a comment

5c5c580

⅄ 22-11-22-inaccessible-case → 22-11-21-bump-lts

20606e7

update stack.yaml.lock

8c149e1

⅄ trunk → 22-11-21-bump-lts

6d6be45

bump to lts-20.0

3076d15

aryairani previously approved these changes Dec 1, 2022

View reviewed changes

aryairani reviewed Dec 1, 2022

View reviewed changes

unison-hashing-v2/src/Unison/Hashing/V2/Term.hs Show resolved Hide resolved

aryairani self-requested a review December 1, 2022 12:38

swap haskeline dep for one that works on windows (:crossed-fingers:)

20dacc9

mitchellwrosen closed this Apr 18, 2023

aryairani reopened this May 19, 2023

aryairani mentioned this pull request May 20, 2023

building on lts-20.21 (ghc 9.2.6) #4008

Closed

1 task

aryairani closed this May 20, 2023

aryairani mentioned this pull request May 24, 2023

building on lts-20.22 (ghc 9.2.7) #4016

Merged

1 task

mitchellwrosen deleted the 22-11-21-bump-lts branch May 15, 2024 22:19

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Chore: bump to lts-20.0 #3642

Chore: bump to lts-20.0 #3642

mitchellwrosen commented Nov 29, 2022

aryairani commented Dec 1, 2022

aryairani commented Dec 1, 2022 •

edited

Loading

aryairani Dec 1, 2022 •

edited

Loading

aryairani Dec 1, 2022 •

edited

Loading

mitchellwrosen Dec 1, 2022

aryairani Dec 1, 2022

mitchellwrosen Dec 1, 2022

mitchellwrosen Dec 2, 2022

aryairani Dec 2, 2022

aryairani Dec 23, 2022

aryairani Dec 1, 2022

mitchellwrosen Dec 1, 2022

mitchellwrosen Dec 2, 2022

aryairani Dec 2, 2022

ChrisPenner commented Dec 2, 2022 •

edited

Loading

mitchellwrosen commented Dec 13, 2022

aryairani commented Dec 23, 2022

ChrisPenner commented Jan 3, 2023 •

edited

Loading

ceedubs commented Jan 4, 2023

mitchellwrosen commented Apr 18, 2023

ChrisPenner commented Apr 18, 2023

aryairani commented May 19, 2023

aryairani commented May 19, 2023

aryairani commented May 20, 2023

Chore: bump to lts-20.0 #3642

Chore: bump to lts-20.0 #3642

Conversation

mitchellwrosen commented Nov 29, 2022

Overview

aryairani commented Dec 1, 2022

aryairani commented Dec 1, 2022 • edited Loading

aryairani Dec 1, 2022 • edited Loading

Choose a reason for hiding this comment

aryairani Dec 1, 2022 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ChrisPenner commented Dec 2, 2022 • edited Loading

mitchellwrosen commented Dec 13, 2022

aryairani commented Dec 23, 2022

ChrisPenner commented Jan 3, 2023 • edited Loading

ceedubs commented Jan 4, 2023

mitchellwrosen commented Apr 18, 2023

ChrisPenner commented Apr 18, 2023

aryairani commented May 19, 2023

aryairani commented May 19, 2023

aryairani commented May 20, 2023

aryairani commented Dec 1, 2022 •

edited

Loading

aryairani Dec 1, 2022 •

edited

Loading

aryairani Dec 1, 2022 •

edited

Loading

ChrisPenner commented Dec 2, 2022 •

edited

Loading

ChrisPenner commented Jan 3, 2023 •

edited

Loading