On 2015-08-27 at 18:38:46 +0200, Simon Peyton Jones wrote:

Great. Is there a ticket? If not, we'll probably lose track of it.

https://ghc.haskell.org/trac/ghc/ticket/10751 :-)

I'm going by my rather poor memory for this. Frankly, I don't really care where the option sits, as long as I don't need a separate build of GHC to avoid LGPL. ________________________________________ From: Ben Gamari Sent: Friday, August 28, 2015 5:48 AM To: Elliot Cameron; GHC developers Cc: Herbert Valerio Riedel Subject: Re: Planning for the 7.12 release Elliot Cameron writes:

I can't seem to find the exact trac ticket, but the ability to swap out "Integer" implementations at link time would be a huge relief on Windows, which suffers from various problems with dynamic linking. I believe it was originally slated for 7.12. Can someone find it? Here's what I did find: https://ghc.haskell.org/trac/ghc/wiki/ReplacingGMPNotes

A related discussion: https://github.com/commercialhaskell/stack/issues/399

As it stands, we've been trying hard to find good ways to provide custom GHC variants more easily to end users.

Hmm, interesting. I'm not sure how realistic it is to make this a link-time option, however, considering that we may inline bindings from whatever integer package we compile against into the user's program. Herbert, do you have any thoughts on this? Cheers, - Ben

For the record, I wanted to implement this feature a couple of months ago but then got side-tracked. If somebody wants to pick it up (which would be great), please lemme know. In the meantime I've created https://ghc.haskell.org/trac/ghc/ticket/10803 and write up the little information I have on this topic already On 2015-08-28 at 09:20:54 +0200, Simon Peyton Jones wrote:

Lennart always says he’s going to work on it, but he’s a busy man and nothing has actually happened.

It’s a pretty easy feature to implement I think.

Simon

From: ghc-devs [mailto:ghc-devs-bounces@haskell.org] On Behalf Of Sean Leather Sent: 28 August 2015 06:40 To: Ben Gamari Cc: GHC developers Subject: Re: Planning for the 7.12 release

On Thu, Aug 27, 2015 at 5:38 PM, Ben Gamari wrote: These items are a bit less certain but may make it in if the authors push forward quickly enough,

* Support for Type Signature Sections, allowing you to write (:: ty) as a shorthand for (\x -> x :: ty).

Once Lennart convinced me of the usefulness of this [1], I've been finding plenty of places where I wish I had it. Who's working on it? What's the status? Is there a ticket for it?

Regards, Sean

[1] http://augustss.blogspot.com/2014/04/a-small-haskell-extension.html _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs

-- "Elegance is not optional" -- Richard O'Keefe

Andrey Mokhov is the man to ask. I'm copying him. Simon | -----Original Message----- | From: ghc-devs [mailto:ghc-devs-bounces@haskell.org] On Behalf Of | Kosyrev Serge | Sent: 28 August 2015 11:31 | To: Ben Gamari | Cc: GHC developers | Subject: Re: Planning for the 7.12 release | | Ben Gamari writes: | > These items are a bit less certain but may make it in if the authors | > push forward quickly enough, | | [..] | | > * A (possible) overhaul of GHC's build system to use Shake | instead | > of Make. | | Is there a breakdown of what remains to be done on this front? | | -- | с уважениeм / respectfully, | Косырев Серёга | _______________________________________________ | ghc-devs mailing list | ghc-devs@haskell.org | http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs

Kosyrev Serge <_deepfire@feelingofgreen.ru> writes:

Ben Gamari writes:

...

These items are a bit less certain but may make it in if the authors push forward quickly enough,

[..]

...

* A (possible) overhaul of GHC's build system to use Shake instead of Make.

Is there a breakdown of what remains to be done on this front?

I'm actually not entirely clear on the general plan regarding the Shake-up and perhaps this is a good time to discuss it. How gradual of a transition do we envision this will be? Specifically, for how long do we want the two build systems to coexist (if at all)? If this is intended to be an immediate wholesale switch to Shake I would be very skeptical of merging for 7.12 as three months is, in my opinion, very little time to test such a sweeping change. However, a long transition time is also not terribly desirable as maintaining two largely independent build systems potentially carries significant cost. Can we in principle expect the Shake build system to build all of the configurations GHC currently supports from day-one (including, for instance, cross compilation)? The batch files in the repository suggest that it has been used on Windows but has it been tested on our other Tier 1 platforms? I just attempted to use the current state of the repository and sadly found that things fell apart pretty quickly [1]. I would love to see this happen, but obviously we need to tread carefully when performing such a major overhaul to code so central to the project. Moreover, I would really like to minimize the probability that we increase the maintenance burden of our build infrastructure. I think if there is clear communication regarding what remains to be done and motivation to quickly finish these items then Shakification is still an possibility for 7.12. Otherwise I personally think we may want to be a bit conservative. End users can always check out the shaking-up-ghc repository into their GHC trees themselves if they want to try using it. I'm certainly willing to consider other opinions, however. Cheers, - Ben [1] After I manually ran ./boot, $ _shake/build --lint --directory ".." $@ ... # various output from ./configure Reading shake/cfg/system.config... Reading package dependencies... Error when running Shake build system: * OracleQ (PackageDataKey ("libraries/bin-package-db/dist-boot/package-data.mk","libraries_bin-package-db_dist-boot_LIB_NAME")) * libraries/bin-package-db/dist-boot/package-data.mk * libraries/bin-package-db/dist-boot/package-data.mk libraries/bin-package-db/dist-boot/haddock-prologue.txt libraries/bin-package-db/dist-boot/inplace-pkg-config libraries/bin-package-db/dist-boot/setup-config libraries/bin-package-db/dist-boot/build/autogen/cabal_macros.h * libraries/binary/dist-boot/package-data.mk * libraries/binary/dist-boot/package-data.mk libraries/binary/dist-boot/haddock-prologue.txt libraries/binary/dist-boot/inplace-pkg-config libraries/binary/dist-boot/setup-config libraries/binary/dist-boot/build/autogen/cabal_macros.h * /mnt/work/ghc/ghc-shake/inplace/bin/ghc-cabal Error, file does not exist and no rule available: /mnt/work/ghc/ghc-shake/inplace/bin/ghc-cabal

| If this is intended to be an immediate wholesale switch to Shake I | would be very skeptical of merging for 7.12 as three months is, in my | opinion, very little time to test such a sweeping change. No there is no chance that we switch over to Shake for 7.12. At the moment it's nowhere near ready for prime time. But I do hope that we'll have material progress on a side-by-side build system before 7.12 comes out. It probably won't do everything but it should do some things well. I hope. Simon | | However, a long transition time is also not terribly desirable as | maintaining two largely independent build systems potentially carries | significant cost. | | Can we in principle expect the Shake build system to build all of the | configurations GHC currently supports from day-one (including, for | instance, cross compilation)? The batch files in the repository | suggest that it has been used on Windows but has it been tested on our | other Tier 1 platforms? | | I just attempted to use the current state of the repository and sadly | found that things fell apart pretty quickly [1]. | | I would love to see this happen, but obviously we need to tread | carefully when performing such a major overhaul to code so central to | the project. Moreover, I would really like to minimize the probability | that we increase the maintenance burden of our build infrastructure. | | I think if there is clear communication regarding what remains to be | done and motivation to quickly finish these items then Shakification | is still an possibility for 7.12. | | Otherwise I personally think we may want to be a bit conservative. End | users can always check out the shaking-up-ghc repository into their | GHC trees themselves if they want to try using it. I'm certainly | willing to consider other opinions, however. | | Cheers, | | - Ben | | | [1] After I manually ran ./boot, | | $ _shake/build --lint --directory ".." $@ | ... # various output from ./configure | Reading shake/cfg/system.config... | Reading package dependencies... | Error when running Shake build system: | * OracleQ (PackageDataKey ("libraries/bin-package-db/dist- | boot/package-data.mk","libraries_bin-package-db_dist-boot_LIB_NAME")) | * libraries/bin-package-db/dist-boot/package-data.mk | * libraries/bin-package-db/dist-boot/package-data.mk | libraries/bin-package-db/dist-boot/haddock-prologue.txt libraries/bin- | package-db/dist-boot/inplace-pkg-config libraries/bin-package-db/dist- | boot/setup-config libraries/bin-package-db/dist- | boot/build/autogen/cabal_macros.h | * libraries/binary/dist-boot/package-data.mk | * libraries/binary/dist-boot/package-data.mk | libraries/binary/dist-boot/haddock-prologue.txt libraries/binary/dist- | boot/inplace-pkg-config libraries/binary/dist-boot/setup-config | libraries/binary/dist-boot/build/autogen/cabal_macros.h | * /mnt/work/ghc/ghc-shake/inplace/bin/ghc-cabal | Error, file does not exist and no rule available: | /mnt/work/ghc/ghc-shake/inplace/bin/ghc-cabal

Simon Peyton Jones writes:

Actually that’s a good idea.

Simon

From: ghc-devs [mailto:ghc-devs-bounces@haskell.org] On Behalf Of Greg Weber Sent: 28 August 2015 16:43 To: Ben Gamari Cc: GHC developers Subject: Re: Planning for the 7.12 release

Can we call this GHC 8.0 instead of 7.12 ? Overloaded record fields and backtraces are a huge missing piece to Haskell. It would be nice to have the bump to celebrate this occasion and say that Haskell 8 is "ready". I have had a hard time seriously recommending Haskell due to those last missing features. Now I should be able to say without reservation: "use Haskell > 8; it is great!"

I was discussing this very matter yesterday with a few folks. I think we certainly have enough features in this release to do a major bump. I half-jokingly suggested that 8.0 should only come with Phase 2 of Richard's Dependent Haskell work, but I'm willing to settle for merely kind equality. I think doing a major bump would be a great idea. Cheers, - Ben

On Aug 28, 2015, at 1:33 PM, Ben Gamari wrote:

I half-jokingly suggested that 8.0 should only come with Phase 2 of Richard's Dependent Haskell work, but I'm willing to settle for merely kind equality.

I think doing a major bump would be a great idea.

Drat! I, too, was hoping to herald in 8.0 with -XDependentTypes, but I guess I'm a little late. Don't hold up this change for me, though. :) Richard