May 2016 - ghc-devs - Haskell.org

instances for closed type families
by Alan & Kim Zimmerman 26 May '16

26 May '16

I am working on https://ghc.haskell.org/trac/ghc/ticket/12105, and have a type to ensure that the `HsMatchContext` has either a `RdrName` or a `Name`, not an `Id`. type family NameOrRdrName id where NameOrRdrName Id = Name NameOrRdrName Name = Name NameOrRdrName RdrName = RdrName Is there any way to declare `Data` and `OutputableBndr` instances for this? Without it I am having to do something like instance (OutputableBndr name, OutputableBndr (NameOrRdrName name)) => Outputable (HsDecl name) where which requires UndecidableInstances. I get compiler/hsSyn/PlaceHolder.hs:114:19: Illegal type synonym family application in instance: NameOrRdrName id In the stand-alone deriving instance for ‘Data (NameOrRdrName id)’ for the parameterised version and compiler/hsSyn/PlaceHolder.hs:115:19: Illegal type synonym family application in instance: NameOrRdrName RdrName In the stand-alone deriving instance for ‘Data (NameOrRdrName RdrName)’ for the one specific to `RdrName`. Regards Alan

2 2

Build fails
by Simon Peyton Jones 24 May '16

24 May '16

Binary.readBinMem: decompression failed What’s going on? This is a clean validate on Linux. I have done git pull, git submodule update –recursive –init Simon cd libraries && sh gen_contents_index --intree grep: ../bin-package-db/bin-package-db.cabal: No such file or directory grep: ../integer-gmp2/integer-gmp.cabal: No such file or directory haddock: internal error: Binary.readBinMem: decompression failed CallStack (from HasCallStack): error, called at compiler/utils/Binary.hs:192:16 in ghc:Binary make[1]: *** [libraries/dist-haddock/index.html] Error 1 make: *** [all] Error 2

2 5

Re: [Haskell-cafe] [ANNOUNCE] GHC 8.0.1 is available!
by Ben Gamari 22 May '16

22 May '16

KC <kc1956(a)gmail.com> writes: > Hopefully this makes it into the Haskell Platform soon > I believe a Haskell Platform release is imminent, not sure about Stackage. Cheers, - Ben

1 0

HEAD failing
by Simon Peyton Jones 19 May '16

19 May '16

Devs, I’m getting this on a clean build of HEAD on Linux Unexpected failures: /tmp/ghctest/0MzM7D/1/2/3/./rts/T11108 T11108 [bad stdout] (normal) /tmp/ghctest/0MzM7D/1/2/3/./rts/T8308/T8308 T8308 [bad stdout] (normal) Two things · would it be possible to remove the long /tmp/blah prefix? · What’s wrong with these two rts tests? I see that #11108 already reports the failure, and I’ve re-opened #8308 Simon

2 3

8.0.1 source tarballs available (yet again)
by Ben Gamari 19 May '16

19 May '16

tl;dr: We are giving this release another attempt. Cross your fingers and try building again. Hello GHC packagers, Thanks for your help in the last few days working through what were hopefully the last issues in the release. I've uploaded a new set of source tarballs at, http://downloads.haskell.org/~ghc/8.0.1/ These tarballs are derived from, 787900138b94790ddd13d76c2853d789d3335a5d This commit differs from the previous tarball in a number of ways, * a variety of packaging issues surrounding haddock's documentation have been fixed * the ghc-boot library has been split up. It was pointed out quite late that the previous structure of ghc-boot greatly increased the dependency footprint of the template-haskell package, which would have made it nearly impossible to upgrade the bytestring and binary packages. * a pair of fixes for crashes on PowerPC have been merged * a patch addressing a serious performance issue in parallel garbage collection has been merged * the bytestring module has been bumped to 0.8.10.1, fixing a rather serious correctness issue on big-endian systems We'll have the usual one-week build window for binary distributions. As always, let either Austin or I know if you have any trouble building your distribution. I have yet to push the ghc-8.0.1 tag in case we encounter unexpected issues but all of my builds with this tarball thusfar have gone quite well (then again, so did those of the previous tarball). Good luck and thanks for all of your work! Cheers, - Ben

6 13

Re: [GHC] #11140: add command-line option to GHC to dump raw parse trees of Haskell programs
by Alan & Kim Zimmerman 18 May '16

18 May '16

I'm with mpickering on this one. Each stage has a debug format that makes sense for the stage. For parsing it is the exact ast Alan On 18 May 2016 12:43 PM, "GHC" <ghc-devs(a)haskell.org> wrote: > #11140: add command-line option to GHC to dump raw parse trees of Haskell > programs > -------------------------------------+------------------------------------- > Reporter: bollu | Owner: > Type: feature request | Status: new > Priority: low | Milestone: > Component: Compiler | Version: 7.10.2 > Resolution: | Keywords: > Operating System: Unknown/Multiple | Architecture: > | Unknown/Multiple > Type of failure: None/Unknown | Test Case: > Blocked By: | Blocking: > Related Tickets: | Differential Rev(s): > Wiki Page: | > -------------------------------------+------------------------------------- > > Comment (by mpickering): > > That is true but by them stages in the compilation pipeline you don't care > about what the raw format looks like. When modifying the parser you need > to know the precise structure of the AST rather than some pretty printed > version produced by `Outputtable`. > > -- > Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11140#comment:6> > GHC <http://www.haskell.org/ghc/> > The Glasgow Haskell Compiler >

1 0

HEADS UP: running tests in /tmp and the `extra_files` setup function
by Thomas Miedema 18 May '16

18 May '16

Hello GHC developers, the testsuite driver now runs each test in a temporary directory in /tmp, after first linking/copying all files that the test requires to that directory. When adding a new test: ** if your test requires source files that don't start with the name of your test, then you *have to* specify those files using the `extra_files` setup function. [1]* * you no longer have to specify files to cleanup (`clean_files` and `clean_cmd` are deprecated) * you no longer have to add generated files to `testsuite/.gitignore` * you no longer have to worry about two tests possibly overwriting each others intermediate (.o, .hi) files, and you no longer have to specify `-outputdir` if you want to let multiple tests use the same source files. Example: *BEFORE:* test('driver011', extra_clean(['A011.hi', 'A011.o']), run_command, ['$MAKE -s --no-print-directory test011']) *AFTER:* test('driver011', extra_files(['A011.hs']), run_command, ['$MAKE -s --no-print-directory test011']) See https://ghc.haskell.org/trac/ghc/ticket/11980 and https://ghc.haskell.org/trac/ghc/wiki/Building/RunningTests/Adding for more information. Cheers, Thomas [1] Extra files for existing tests are currently listed in `testsuite/driver/extra_files.py`.

3 3

New wiki page about tying the knot in GHC
by Edward Z. Yang 18 May '16

18 May '16

Hello all, I've written a new wiki page about how knot tying works in GHC: https://ghc.haskell.org/trac/ghc/wiki/Commentary/Compiler/TyingTheKnot Please let me know if anything is unclear and I'll try to revise it for more clarity. Edward

2 2

Phabricator not sending emails
by Edward Z. Yang 17 May '16

17 May '16

As far as I can tell, Phabricator is not sending emails. I noticed when some of my diffs updated but I didn't get any emails, and doublechecked by asking for a password reset email. Edward

2 1

Added submodule to haddock
by Ben Gamari 17 May '16

17 May '16

Hello everyone, Note that a submodule has been added to Haddock in a recent merge. You may need to check it out with (in the root of the GHC source tree), git submodule update --init --recursive to ensure it's available. Cheers, - Ben

1 0