pkgsMusl.haskell.compiler.ghc8107Binary: Remove now-incorrect gmp dependency #138523

nh2 · 2021-09-19T12:24:56Z

Motivation for this change

GHC HQ switched the musl bindists from gmp to integer-simple
with GHC >= 8.10.6, but this was not reflected in the nixpkgs update:

commit 6f12424: ghc: 8.10.5-binary -> 8.10.7-binary
From PR haskellPackages: update stackage and hackage #135453

Things done

domenkozar · 2021-09-19T13:44:16Z

I had no idea they switched.

cdepillabout · 2021-09-19T15:07:55Z

@nh2 How do you recommend this be tested?

Would building the following things fully test this?

binary ghcs

haskell.packages.ghc8102Binary.lens: This would test that the gmp-linked non-musl ghc8102Binary can actually be used to build packages.
pkgsMusl.haskell.packages.ghc8102Binary.lens: This would test that the gmp-linked musl ghc8102Binary can be used to build packages.
haskell.packages.ghc8107Binary.lens: This would test that the gmp-linked non-musl ghc8107Binary can actually be used to build packages.
pkgsMusl.haskell.packages.ghc8107Binary.lens: This would test that the non-gmp-linked musl ghc8107Binary can be used to build packages.

ghc8107

haskell.packages.ghc8107.lens: This would test that a gmp-linked non-musl ghc8107 (which is built with the gmp-linked non-musl ghc865Binary) can be used to build packages. edit: I guess this shouldn't actually be affected by this PR at all
pkgsMusl.haskell.packages.ghc8107.lens: This would test that a gmp-linked musl ghc8107 (which is built with gmp-linked musl ghc8102BinaryMinimal) can be used to build packages.
haskell.packages.integer-simple.ghc8107.lens: This would test that a non-gmp-linked non-musl ghc8107 (which is built with the gmp-linked non-musl ghc865Binary) can be used to build packages. edit: I guess this shouldn't actually be affected by this PR at all
pkgsMusl.haskell.packages.integer-simple.ghc8107.lens: This would test that a non-gmp-linked musl ghc8107 (which is built with gmp-linked musl ghc8102BinaryMinimal) can be used to build packages.

ghc901

haskell.packages.ghc901.lens: This would test that a gmp-linked non-musl ghc901 (which is built with the gmp-linked non-musl ghc8102Binary) can be used to build packages.
pkgsMusl.haskell.packages.ghc901.lens: This would test that a gmp-linked musl ghc901 (which is built with gmp-linked musl ghc8102Binary) can be used to build packages.
haskell.packages.integer-simple.ghc901.lens: This would test that a non-gmp-linked non-musl ghc901 (which is built with the gmp-linked non-musl ghc8102Binary) can be used to build packages.
pkgsMusl.haskell.packages.integer-simple.ghc901.lens: This would test that a non-gmp-linked musl ghc901 (which is built with gmp-linked musl ghc8102Binary) can be used to build packages.

Is this about right? Do the success of these derivations give a good indication as to whether this PR is correct? Or is this overkill?

The one thing the above derivation list is missing is something like the above ghc901 section, but where haskell.compiler.ghc901, pkgsMusl.haskell.compiler.ghc901, haskell.compiler.integer-simple.ghc901, pkgsMusl.haskell.compiler.integer-simple.ghc901 are built using ghc8107Binary.

Also, in pkgs/top-level/haskell-packages.nix, there is a note that starting from GHC 9, integer-simple and integer-gmp have been replaced by ghc-bignum with native and gmp backends. Should these be tested as well??

nh2 · 2021-09-19T21:39:51Z

@cdepillabout Personally I'd just build all Haskell packages Hydra builds, and check if any of them regressed, if that's easy (I fail to remember how to do that though). However, all the packages you mentioned build fine for me (see below).

That said, I think it'd be better to test application builds than library ones, given that we were dealing with linker errors.

Recording here some failures and successes:

Failure: pkgsMusl.haskell.compiler.integer-simple.ghc884

/build/ghc27986_0/ghc_4.c:7:15: error:
     error: variable ‘stg_exports_FastString’ has initializer but incomplete type
        7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
          |               ^~~~~~~~~~~~~~~~~~
  |
7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
  |               ^

/build/ghc27986_0/ghc_4.c:7:61: error:
     error: ‘struct ForeignExportsList’ has no member named ‘exports’
        7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
          |                                                             ^~~~~~~
  |
7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
  |                                                             ^

/build/ghc27986_0/ghc_4.c:7:72: error:
     error: extra brace group at end of initializer
        7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
          |                                                                        ^
  |
7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
  |                                                                        ^

Success: pkgsMusl.haskellPackages.hello
Success: pkgsMusl.haskell.compiler.integer-simple.ghc8107
Success: haskell.packages.ghc8102Binary.lens
Success: pkgsMusl.haskell.packages.ghc8102Binary.lens
Success: haskell.packages.ghc8107Binary.lens
Success: pkgsMusl.haskell.packages.ghc8107Binary.lens
Success: haskell.packages.ghc8107.lens
Success: pkgsMusl.haskell.packages.ghc8107.lens
Success: haskell.packages.integer-simple.ghc8107.lens
Success: pkgsMusl.haskell.packages.integer-simple.ghc8107.lens
Success: haskell.packages.ghc901.lens
Success: pkgsMusl.haskell.packages.ghc901.lens
Success: haskell.packages.integer-simple.ghc901.lens
Success: pkgsMusl.haskell.packages.integer-simple.ghc901.lens

sternenseemann · 2021-09-20T15:57:26Z

pkgs/development/compilers/ghc/8.10.2-binary.nix

@@ -140,20 +143,35 @@ stdenv.mkDerivation rec {

  src = fetchurl binDistUsed.src;

+  # Note that for GHC 8.10 versions <= 8.10.5, the GHC HQ musl bindist


I think this sentence is missing a word.

sternenseemann · 2021-09-20T16:02:06Z

pkgs/development/compilers/ghc/8.10.7-binary.nix

  nativeBuildInputs = [ perl ];
  propagatedBuildInputs =
    lib.optionals useLLVM [ llvmPackages.llvm ]
-    # Because musl bindists currently provide no way to tell where
-    # libgmp is (see not [musl bindists have no .buildinfo]), we need


Can the comment on musl bindists have no .buildinfo further down removed here as well (or clarified wrt musl)?

Thanks, I have fixed the dangling reference.

sternenseemann · 2021-09-20T16:05:29Z

Maybe let's close this in favor of #138524 for simplicity's sake? Would like to get all changes in this rotation.

…endency. GHC HQ switched the musl bindists from gmp to `integer-simple` with GHC >= 8.10.6, but this was not reflected in the nixpkgs update: * commit 6f12424: ghc: 8.10.5-binary -> 8.10.7-binary From PR NixOS#135453 See also NixOS#130441.

nh2 · 2021-09-20T17:04:15Z

Maybe let's close this in favor of #138524 for simplicity's sake? Would like to get all changes in this rotation.

Sure, can do either.

I made this one separately because it is much easier, as it touches only musl; the other one touches all compilers.

sternenseemann · 2021-09-20T17:06:45Z

I'd just do the set rebuild all in one go, actually.

nh2 · 2021-09-26T17:18:17Z

Failure: pkgsMusl.haskell.compiler.integer-simple.ghc884

/build/ghc27986_0/ghc_4.c:7:15: error:
     error: variable ‘stg_exports_FastString’ has initializer but incomplete type
        7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
          |               ^~~~~~~~~~~~~~~~~~
  |
7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
  |               ^

/build/ghc27986_0/ghc_4.c:7:61: error:
     error: ‘struct ForeignExportsList’ has no member named ‘exports’
        7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
          |                                                             ^~~~~~~
  |
7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
  |                                                             ^

/build/ghc27986_0/ghc_4.c:7:72: error:
     error: extra brace group at end of initializer
        7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
          |                                                                        ^
  |
7 | static struct ForeignExportsList stg_exports_FastString = {.exports =  {}, .n_entries =  0};
  |                                                                        ^

@sternenseemann Investigating this deeper, I found some odd results:

let mc = pkgsMusl.haskell.compiler in:

ghc8102BinaryMinimal succeeds at mc.ghc884 fine.
ghc8107BinaryMinimal fails at mc.ghc884, with the above C error,
so the .2 -> .7 bindist regressed that
ghc8102BinaryMinimal fails at mc.integer-simple.ghc884, with the above C error
ghc8102BinaryMinimal fails at mc.integer-simple.ghc884 with the old expected error
ld: cannot find -lgmp

Reverse bootstrapping is not supported by GHC upstream. In the case of 8.8.4 it just happens to work using 8.10.2, with later versions, specifically 8.10.7 there seems to be some digressions in the generated / used C code which cause 8.8.4 to fail to compile [1]. Thus we revert to using 8.10.2 for aarch64 and Musl which means: Still no integer-simple and musl at the same time (however all other GHCs have it, so it's probably not a problem) and no aarch64-darwin (GHC 8.8.4 can't target that architecture anyways). In short, the situation stays the same. [1]: #138523 (comment)

nh2 requested review from cdepillabout, maralorn and sternenseemann September 19, 2021 12:24

nh2 requested a review from expipiplus1 as a code owner September 19, 2021 12:24

github-actions bot added the 6.topic: haskell label Sep 19, 2021

This was referenced Sep 19, 2021

ghc: mark broken for musl + integer-simple builds #130441

Merged

ghc: Add variantSuffix, fix musl+integer-simple #138524

Closed

nh2 force-pushed the ghc-binary-musl-propagated-comment branch from 13744ad to fb79cf9 Compare September 19, 2021 13:38

cdepillabout approved these changes Sep 20, 2021

View reviewed changes

sternenseemann reviewed Sep 20, 2021

View reviewed changes

nh2 force-pushed the ghc-binary-musl-propagated-comment branch from fb79cf9 to a51af7d Compare September 20, 2021 17:02

ofborg bot requested review from cdepillabout, domenkozar, prusnak, guibou and sternenseemann September 20, 2021 21:12

ofborg bot added 10.rebuild-darwin: 0 This PR does not cause any packages to rebuild on Darwin 10.rebuild-linux: 11-100 labels Sep 20, 2021

sternenseemann merged commit 14d020b into NixOS:haskell-updates Sep 23, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

pkgsMusl.haskell.compiler.ghc8107Binary: Remove now-incorrect gmp dependency #138523

pkgsMusl.haskell.compiler.ghc8107Binary: Remove now-incorrect gmp dependency #138523

nh2 commented Sep 19, 2021 •

edited

Loading

domenkozar commented Sep 19, 2021

cdepillabout commented Sep 19, 2021 •

edited

Loading

nh2 commented Sep 19, 2021 •

edited

Loading

sternenseemann Sep 20, 2021

nh2 Sep 20, 2021

sternenseemann Sep 20, 2021

nh2 Sep 20, 2021

sternenseemann commented Sep 20, 2021 •

edited

Loading

nh2 commented Sep 20, 2021

sternenseemann commented Sep 20, 2021

nh2 commented Sep 26, 2021

		@@ -140,20 +143,35 @@ stdenv.mkDerivation rec {

		src = fetchurl binDistUsed.src;

		# Note that for GHC 8.10 versions <= 8.10.5, the GHC HQ musl bindist

pkgsMusl.haskell.compiler.ghc8107Binary: Remove now-incorrect gmp dependency #138523

pkgsMusl.haskell.compiler.ghc8107Binary: Remove now-incorrect gmp dependency #138523

Conversation

nh2 commented Sep 19, 2021 • edited Loading

Motivation for this change

Things done

domenkozar commented Sep 19, 2021

cdepillabout commented Sep 19, 2021 • edited Loading

nh2 commented Sep 19, 2021 • edited Loading

sternenseemann Sep 20, 2021

Choose a reason for hiding this comment

nh2 Sep 20, 2021

Choose a reason for hiding this comment

sternenseemann Sep 20, 2021

Choose a reason for hiding this comment

nh2 Sep 20, 2021

Choose a reason for hiding this comment

sternenseemann commented Sep 20, 2021 • edited Loading

nh2 commented Sep 20, 2021

sternenseemann commented Sep 20, 2021

nh2 commented Sep 26, 2021

nh2 commented Sep 19, 2021 •

edited

Loading

cdepillabout commented Sep 19, 2021 •

edited

Loading

nh2 commented Sep 19, 2021 •

edited

Loading

sternenseemann commented Sep 20, 2021 •

edited

Loading