haskell-language-server, stylish-haskell: Fix build on GHC 8.10.* #97744

Galagora · 2020-09-11T08:33:49Z

Motivation for this change

Haskell Language Server uses a fork of brittany that compiles on GHC 8.10.*. In addition, stylish-haskell forgot to update their Cabal version bounds in the test suite, causing a hard-to-debug build failure.

Closes #97743.
Closes #97745.

Things done

cdepillabout

Could you change this to target the haskell-updates branch instead of master?

Other than the branch and the the one suggestion below, this looks good to me!

cc @maralorn and @GuillaumeDesforges

pkgs/development/haskell-modules/configuration-common.nix

GuillaumeDesforges

LGTM

GuillaumeDesforges · 2020-09-11T09:49:59Z

pkgs/development/tools/haskell/haskell-language-server/hls-ghcide.nix

 }:
 mkDerivation {
  pname = "ghcide";
-  version = "0.2.0";
+  version = "0.3.0";


Not sure if that's an issue in itself, but having a single update.sh makes fixing or extending with other libs (like this PR) into one non-atomic commit. While nixpkgs is a bit blurry about its "atomic changes" policy, I think it is still a good idea to split package update and build fixes.

@cdepillabout do you consider it to not be an issue?
Should this PR be split into more atomic commits manually? (my preference)
Do we change the workflow by splitting the update scripts?
Anything else?

I split out the commits and added a note to update.sh to remind others to do the same. I don't think the script should be split out; you'll probably want to test the binary with all the automated updates applied, and you'll probably want to commit only after the binary works.

I don't particularly worry about it that much. I agree that it is a good idea to have atomic changes, but there is some trade-off with making it unreasonably challenging for people to contribute.

If you don't split out your commits, I guess you run the risk of having a whole bunch of unrelated changes reverted if something in one of your commits needs to be reverted.

I would be fine with the HLS stuff all going into one commit, since it is all basically related anyway.

Galagora · 2020-09-11T11:47:21Z

Sorry, that was a mistake...

maralorn

LGTM

If this builds, this should be fine. Thanks!

maralorn · 2020-09-11T12:06:05Z

@GrahamcOfBorg build haskellPackages.haskell-language-server
@GrahamcOfBorg build haskellPackages.stylish-haskell
@GrahamcOfBorg build haskell.packages.ghc8102.haskell-language-server
@GrahamcOfBorg build haskell.packages.ghc8102.hls-brittany

cdepillabout · 2020-09-11T12:48:27Z

@GrahamcOfBorg build haskellPackages.haskell-language-server
@GrahamcOfBorg build haskellPackages.stylish-haskell
@GrahamcOfBorg build haskell.packages.ghc8102.haskell-language-server
@GrahamcOfBorg build haskell.packages.ghc8102.hls-brittany

maralorn · 2020-09-11T16:19:23Z

Can you have another look at those tests? There seems to be a wrong hash error going on?

Galagora · 2020-09-11T16:58:29Z

I can't reproduce the stylish-haskell patch hash error.

$ nix-prefetch-url https://github.com/jaspervdj/stylish-haskell/commit/9550aa1cd177aa6fe271d075177109d66a79e67f.patch
[0.0 MiB DL]
path is '/nix/store/k82rvk4aid72dvn72v2qqr5vdhv8amwd-9550aa1cd177aa6fe271d075177109d66a79e67f.patch'
159jr80k40hdq5gpqfjknqx6vj2licx1l0f57l5r3k4264lnxjdb

On aarch64 the hash fails to validate, but on x86_64-linux the build just... fails? I can't tell why (maybe the download failed?).

copying path '/nix/store/c9sk1y8gqfgkfjmn2rgldx6rff66a7x7-HsYAML-aeson-0.2.0.0' from 'https://cache.nixos.org'...
cannot build derivation '/nix/store/ir4yj9yk3j9a6s69742fj01j2bq278x1-stylish-haskell-0.11.0.3.drv': 1 dependencies couldn't be built
error: build of '/nix/store/ir4yj9yk3j9a6s69742fj01j2bq278x1-stylish-haskell-0.11.0.3.drv' failed

Galagora · 2020-09-11T17:25:05Z

The failed tests are a problem with the build machines, not with the expressions. On aarch64, the wrong hash is computed for the stylish-haskell patch, presumably because that platform's fetchpatch returns a subtly different result. For the rest of the builds, at some point they just randomly fail, with an error message like this:

cannot build derivation '/nix/store/4x1mgipakbpj8b8qcyy5vx0njfvkq9hq-haskell-language-server-0.4.0.0.drv': 1 dependencies couldn't be built

Sometimes accompanied by this:

error: --- Error --- nix-build
error: --- Error --- nix-daemon

There's no indication as to which dependencies actually fail, however.

Galagora · 2020-09-11T18:18:48Z

@GrahamcOfBorg build haskellPackages.haskell-language-server
@GrahamcOfBorg build haskellPackages.stylish-haskell
@GrahamcOfBorg build haskell.packages.ghc8102.haskell-language-server

Galagora · 2020-09-11T18:44:13Z

Ouch

pbogdan · 2020-09-11T19:14:44Z

@Galagora It looks like on aarch64 #97407 still applies which means there's no working GHC on that platform so you can probably ignore it.

For the hash-mismatch - a hash returned by nix-prefetch-url is generally not suitable for use in fetchpatch which does additional post-processing to normalise the contents. However if you ran it locally nix-prefetch-url would register a fixed-output derivation in your store. When you feed that hash to fetchpatch Nix will see it already has it so won't do any work fetchpatch would normally do and just use it as is. On the builders, which don't have it yet, the patch will be fetched, post-processed, which will produce different output with different hash.

You should be able to reproduce it locally if you delete the result of nix-prefetch-url from your store (nix-store --delete on the result).

The patch is merged into master

0.6.2 -> 0.6.3.2

pbogdan · 2020-09-11T19:29:39Z

There's still a hash mismatch in haskellPackages.stylish-haskell.patches…

cdepillabout · 2020-09-12T00:34:49Z

@pbogdan It looks like peti fixed the stylish-haskell patch in 5f7b5a3, which should be in both the haskell-updates branch and master.

Galagora requested a review from cdepillabout as a code owner September 11, 2020 08:33

ofborg bot added 6.topic: haskell 8.has: package (new) 10.rebuild-darwin: 1-10 10.rebuild-linux: 1-10 labels Sep 11, 2020

cdepillabout reviewed Sep 11, 2020

View reviewed changes

pkgs/development/haskell-modules/configuration-common.nix Show resolved Hide resolved

cdepillabout requested review from maralorn and GuillaumeDesforges September 11, 2020 09:33

GuillaumeDesforges reviewed Sep 11, 2020

View reviewed changes

Galagora force-pushed the hls-ghc-8102 branch 2 times, most recently from c3e7f4b to fb06974 Compare September 11, 2020 11:33

Galagora changed the base branch from master to haskell-updates September 11, 2020 11:46

Galagora requested review from cdepillabout and GuillaumeDesforges September 11, 2020 11:46

maralorn approved these changes Sep 11, 2020

View reviewed changes

Galagora force-pushed the hls-ghc-8102 branch from 6061c5a to 9066102 Compare September 11, 2020 18:15

peti force-pushed the haskell-updates branch 2 times, most recently from bd1911c to 10ec99b Compare September 11, 2020 18:30

galagora added 3 commits September 11, 2020 22:17

haskell-language-server: add brittany fork

e9b791e

haskell-language-server: hls-ghcide 0.3.0 -> 0.4.0

80328d9

The patch is merged into master

haskell-language-server: ghc-exactprint update

da9d7a4

0.6.2 -> 0.6.3.2

galagora added 3 commits September 11, 2020 22:17

haskell-language-server: cb861b878a -> c476992f43

Verified

This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.

GPG key ID: 4AEE18F83AFDEB23
Expired

Verified
Learn about vigilant mode

875e9b9

haskellPackages.stylish-haskell: fix cabal bounds

Verified

This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.

GPG key ID: 4AEE18F83AFDEB23
Expired

Verified
Learn about vigilant mode

9ee19bc

haskell-language-server: don't test fourmolu

Verified

This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.

GPG key ID: 4AEE18F83AFDEB23
Expired

Verified
Learn about vigilant mode

5175a89

ofborg bot added the 2.status: merge conflict label Sep 11, 2020

Galagora force-pushed the hls-ghc-8102 branch from 9066102 to 5175a89 Compare September 11, 2020 19:18

peti merged commit 2076624 into NixOS:haskell-updates Sep 11, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

haskell-language-server, stylish-haskell: Fix build on GHC 8.10.* #97744

haskell-language-server, stylish-haskell: Fix build on GHC 8.10.* #97744

Galagora commented Sep 11, 2020 •

edited

Loading

cdepillabout left a comment

GuillaumeDesforges left a comment

GuillaumeDesforges Sep 11, 2020

Galagora Sep 11, 2020 •

edited

Loading

cdepillabout Sep 11, 2020 •

edited

Loading

Galagora commented Sep 11, 2020

maralorn left a comment

maralorn commented Sep 11, 2020 •

edited

Loading

cdepillabout commented Sep 11, 2020

maralorn commented Sep 11, 2020

Galagora commented Sep 11, 2020 •

edited

Loading

Galagora commented Sep 11, 2020 •

edited

Loading

Galagora commented Sep 11, 2020

Galagora commented Sep 11, 2020

pbogdan commented Sep 11, 2020

pbogdan commented Sep 11, 2020

cdepillabout commented Sep 12, 2020

haskell-language-server, stylish-haskell: Fix build on GHC 8.10.* #97744

haskell-language-server, stylish-haskell: Fix build on GHC 8.10.* #97744

Conversation

Galagora commented Sep 11, 2020 • edited Loading

Motivation for this change

Things done

cdepillabout left a comment

Choose a reason for hiding this comment

GuillaumeDesforges left a comment

Choose a reason for hiding this comment

GuillaumeDesforges Sep 11, 2020

Choose a reason for hiding this comment

Galagora Sep 11, 2020 • edited Loading

Choose a reason for hiding this comment

cdepillabout Sep 11, 2020 • edited Loading

Choose a reason for hiding this comment

Galagora commented Sep 11, 2020

maralorn left a comment

Choose a reason for hiding this comment

maralorn commented Sep 11, 2020 • edited Loading

cdepillabout commented Sep 11, 2020

maralorn commented Sep 11, 2020

Galagora commented Sep 11, 2020 • edited Loading

Galagora commented Sep 11, 2020 • edited Loading

Galagora commented Sep 11, 2020

Galagora commented Sep 11, 2020

pbogdan commented Sep 11, 2020

pbogdan commented Sep 11, 2020

cdepillabout commented Sep 12, 2020

Galagora commented Sep 11, 2020 •

edited

Loading

Galagora Sep 11, 2020 •

edited

Loading

cdepillabout Sep 11, 2020 •

edited

Loading

maralorn commented Sep 11, 2020 •

edited

Loading

Galagora commented Sep 11, 2020 •

edited

Loading

Galagora commented Sep 11, 2020 •

edited

Loading