- User Since
- May 19 2015, 3:52 AM (186 w, 2 d)
Oct 29 2018
Rebased on master
Sep 28 2018
Jul 31 2018
Jul 30 2018
I'm not familiar with this part of GHC but here are a few guesses...
Jul 29 2018
- Use with# in ghc-compact
- Implement with[#] with catch#
Jul 28 2018
- Ensure test fails with current HEAD (add "-threaded")
Jul 25 2018
Ok! No problem, as long as it gets fixed, it's fine!
- Add test
Jun 22 2018
Indeed it is only a cosmetic change but the Mach prefix always seemed weird to me. The only machine-specific literals are MachInt, MachWord and MachNullAddr but two of them have already been renamed. All the other ones are specified independently of the actual machine (Float, Double, Word32, Word64, Label, etc.) and/or are not so close to the machine (Unicode char, UTF8 string, Integer, Natural).
- Create a separate diff for the desugaring bug
Jun 21 2018
Jun 20 2018
Jun 15 2018
- Rebae on master
May 13 2018
May 1 2018
Apr 27 2018
Thanks @dfeuer. The new patch for utils/haddock:
--- a/haddock-api/src/Haddock/Backends/LaTeX.hs +++ b/haddock-api/src/Haddock/Backends/LaTeX.hs @@ -169,16 +169,9 @@ ppLaTeXModule _title odir iface = do
Apr 26 2018
- Switch to Ptr and Int. Other minor changes.
Apr 25 2018
Of course the CI cant' build this patch because it needs the following patch in utils/haddock:
Apr 20 2018
- Comment only
- Use Proxy instead of ambiguous types
- Fix constructor in mkNatural
Apr 19 2018
- Fix desugaring of overflowing literals
- Remove pattern synonyms
- Factorise cvt_integer and cvt_natural into cvt_literal
Mar 9 2018
- Simplify hasCafRefs
- Introduce LitNumType and LitNumber constructor
- Factorize some code
Mar 5 2018
@simonpj If we go this way, I think we should replace MachInt, MachInt64, MachWord, MachWord64, LitInteger and LitNatural with something like:
- Add missing stderr file
- Add test for negative Natural literal warning
- Fix build with integer-simple
- Don't use wordToNatural# in base, otherwise built-in rules may produce Natural literals
Mar 2 2018
- wordToNatural: use it and add a built-in rule to lift constant applications into Natural literals
- Add tests
- Warn on negative Natural literals
- Fix some minor glitches
- Fix some line length warnings
- NOINLINE naturalToInteger
- Fix TH
- Fix Bits instance
Mar 1 2018
- Split and add base:GHC.Maybe
- Fix error in naturalFromInteger rule (was returning a LitInteger)
- Ensure that we avoid using Natural literals in base
- Fix use of Natural literal in Data.Bits Natural instance
- Export Natural stuff in basicKnownKeyNames
- Add some missing imports
I have fixed the build. I hope it's ok if I commandeer the revision in order to update my patch. @Bodigrim don't hesitate to commandeer it back.
Feb 26 2018
- naturalToIntegerName, plusNaturalName, minusNaturalName, timesNaturalName, and mkNaturalName are missing from basicKnownKeyNames in compiler/prelude/PrelNames.hs
- import GHC.Types () is missing from libraries/base/GHC/Exception/Type.hs-boot
Feb 16 2018
Well, I got stuck with build error, quoted above. Are you able to build GHC with my patch?
Feb 15 2018
Oct 28 2017
I wouldn't mind waiting for 8.6. Switching back to pattern guards would be quite a lot of (duplicated) work.
Oct 27 2017
Whoa! You mean that adding the patch to HEAD means that 8.2 crashes when compiling HEAD? That's NOT ok.
Oct 26 2017
@bgamari The tests don't pass because of a bug in GHC 8.2.1. I've had to use a snapshot of HEAD to validate this patch (using --enable-bootstrap-with-devel-snapshot).
Oct 19 2017
I was concerned with file cleanup when I proposed to support in-memory object files: suppose we have some TH producing an object file. We could use addForeignObject but we would have no way to remove the temporary object file. While if GHC only writes the temporary file for the duration of the link, there is nothing to cleanup. The API would be similar to addForeignFile (which takes a String, not a FilePath).
- Minor fix in Note
Oct 18 2017
- Factorize multiplications
- Use better patterns
- Add Flag
Jul 1 2017
Jun 19 2017
Jun 15 2017
I didn't want to discuss too much the details of the new hierarchy beforehand because I wasn't sure I would have the patience to finish this patch anyway ;-).
Jun 14 2017
Some submodules also need to be updated:
Feb 2 2017
Jan 23 2017
- Fix test
It looks good to me. Sorry to have missed that!
Jan 9 2017
It looks good to me. Thanks for the patch.