GHC Weekly News - 2015/02/10
thoughtpolice - 2015-02-10
Hi *,
Welcome! This is the first GHC Weekly news of February 2015. You might be wondering what happened to the last one. Well, your editor was just in New York for the past week attending Compose Conference, making friends and talking a lot about Haskell (luckily we missed a snow storm that may have messed it up quite badly!)
The conference was great. I got to have some interesting discussions about GHC and Haskell with many friendly faces from all around at an incredibly well run conference with a stellar set of people. Many thanks to NY Haskell (organizing), Spotify (hosting space), and to all the speakers for a wonderful time. (And of course, your editor would like to thank his employer Well-Typed for sending him!)
But now, since your author has returned, GHC HQ met back up this week for some discussion, with some regularly scheduled topics. For the most part it was a short meeting this week - our goals are pretty well identified:
GHC HQ and the Core Libraries Committee have posted a survey on the future of the 7.10 prelude and the FTP/BBP discussion. The deadline is February 20th, so please vote if the discussion is of interest to you. Simon Peyton-Jones and Simon Marlow will be making the final decision. https://www.haskell.org/pipermail/haskell-cafe/2015-February/118095.html
It’s likely GHC HQ will do a third 7.10.1 Release Candidate at the very end of February after the votes are included. We missed some patches in RC2 (such as Phab:D347) and incorporated even more bugfixes, so this is worth a test drive by users.
For the most part, things for 7.10 have been going very smoothly other than the debates and a few bugs trickling in - there has not been much ticket activity the past two weeks, so things feel pretty good right now. Austin will mostly be focused on shipping 7.10 and keeping the normal review/patch/triaging cycles going until it’s done. We’re on track to fix all the major bugs we’ve assigned (see milestone:7.10.1).
Since my last post, we’ve also had other random assorted chatter on the mailing lists by the dev team:
In light of a recent large bug in GHC which can be used to derive
unsafeCoerce
, GHC HQ has decided to push back the 7.10 release a bit longer to about March, in order to fix this bug and ferret out the little fallout afterwords. It turns out this isn’t a simple bug to fix, but luckily a fix is being worked on already. https://www.haskell.org/pipermail/ghc-devs/2015-January/008189.htmlLuckily, Iavor has started work on fixing this nasty bug, and had a few questions for the list: https://www.haskell.org/pipermail/ghc-devs/2015-February/008269.html
Iavor Diatchki has raised a new topic about a simpler OverloadedRecordsField proposal. Adam swooped in to address some points about the design. https://www.haskell.org/pipermail/ghc-devs/2015-January/008183.html
Herbert Valerio Riedel posted about a huge (76x) regression between GHC 7.11 and GHC 7.10, but strangely nobody has picked up as to why this is the case yet! https://www.haskell.org/pipermail/ghc-devs/2015-January/008207.html
David Feuer has a question: why is
undefined
so special? In particular, it seems as ifundefined
can be specially used as a value with a type of kind#
as well as*
. It turns out GHC has a special notion of subkinding, andundefined
has a type more special than meets the eye which allows this, as Adam Gundry replied. https://www.haskell.org/pipermail/ghc-devs/2015-February/008222.htmlMerijn Verstraaten has started up a discussion about a new proposal of his, ValidateMonoLiterals. The proposal revolves around the idea of using GHC to enforce compile-time constraints on monomorphic literals, whose type may have invariants enforced on them. While this is doable with Template Haskell, Merijn would like to see something inside GHC instead. https://www.haskell.org/pipermail/ghc-devs/2015-February/008239.html
David Feuer asked: can we merge
FlexibleContexts
withFlexibleInstances
? The proposal seems to be relatively undiscussed at the moment with a neutral future, but perhaps someone would like to chime in on this minor issue. https://www.haskell.org/pipermail/ghc-devs/2015-February/008245.htmlGreg Weber opened up a discussion about ‘Restricted Template Haskell’, which would hopefully make it easier for users to see what a TH computation is actually doing. It turns out - as noted by Simon - that Typed Template Haskell is perhaps closer to what Greg wants. The proposal and discussion then resulted in us realising that the typed TH documentation is rather poor! Hopefully Greg or someone can swing in to improve things. https://www.haskell.org/pipermail/ghc-devs/2015-February/008232.html
Closed tickets the past two weeks include: #10028, #10040, #10031, #9935, #9928, #2615, #10048, #10057, #10054, #10060, #10017, #10038, #9937, #8796, #10030, #9988, #10066, #7425, #7424, #7434, #10041, #2917, #4834, #10004, #10050, #10020, #10036, #9213, and #10047.