- ghc-devs - Haskell.org

Changes in GHC API modularity with the "Encode shape information in PMOracle" MR
by Shayne Fletcher 17 Sep '19

17 Sep '19

Some time back, the `ghc-lib` project split into two targets : `ghc-lib-parser` for those projects that just need to produce syntax trees and `ghc-lib` (re-exporting `ghc-lib-parser` modules) having the remaining modules for projects that need to go on and distill parse trees to Core. The idea of course was to reduce build times for tools like hlint that only need parse trees. Roughly, `ghc-lib-parser` got about 200 files and `ghc-lib` 300. Today after landing `7915afc6bb9539a4534db99aeb6616a6d145918a`, "Encode shape information in `PmOracle`", `ghc-lib-parser` now needs 543 files and `ghc-lib` gets just 25. That may be just bad luck for `ghc-lib-parser` and the way it has to be but I thought I should at least mention the knock-on effect of this change on the modularity of the GHC API in case this consequence hasn't been considered? -- *Shayne Fletcher* Language Engineer */* +1 917 699 7663 *Digital Asset* <https://digitalasset.com/>, creators of *DAML <https://daml.com/>* -- This message, and any attachments, is for the intended recipient(s) only, may contain information that is privileged, confidential and/or proprietary and subject to important terms and conditions available at http://www.digitalasset.com/emaildisclaimer.html <http://www.digitalasset.com/emaildisclaimer.html>. If you are not the intended recipient, please delete this message.

3 8

compiler plugin pass order
by Anton Antich 16 Sep '19

16 Sep '19

Dear all, we started exploring Haskell-to-DotNet compiler in frames of a larger project and would like to clarify one issue: is it possible to control the order in the compilation pipeline in which the compiler plugins get called? Inserting .Net compilation via Plugin seems a cleaner way than using GHC API from scratch for a lot of reasons, but the challenge is as follows: we compile from STG but inside the plugin core2core passes have not been called yet, and without them some important optimizations are missing. To illustrate the issue, this function "plus x = x + 4" gets compiled to a nice stg function with core2core (core2core, corePrep, coreToStg, stg2stg): plus :: forall a. Num a => a -> a [LclIdX, Arity=2, Unf=OtherCon []] = [] \r [$dNum_s1bl x] let { sat_s1bz [Occ=Once] :: a [LclId] = [$dNum_s1bl] \u [] let { sat_s1by [Occ=Once] :: Integer [LclId] = CCCS S#! [4#]; } in fromInteger $dNum_s1bl sat_s1by; } in + $dNum_s1bl x sat_s1bz; Inside the plugin, it is only possible to do the same pipeline but *without* the core2core passes, and it gets compiled with additional THUNK: [sat_s1gp :: forall a. Num a => a -> a [LclId] = [] \r [$dNum_s1ga x] let { sat_s1go [Occ=Once] :: a [LclId] = [$dNum_s1ga] \u [] let { sat_s1gn [Occ=Once] :: Integer [LclId] = CCCS S#! [4#]; } in fromInteger $dNum_s1ga sat_s1gn; } in + $dNum_s1ga x sat_s1go;, plus :: forall a. Num a => a -> a [LclIdX] = [] \u [] sat_s1gp;, So the question is - is there any way to ask GHC to run the plugin as LAST in the core2core passes process so that we get the optimized core and convert it to stg? Alternatively, maybe STG plugins should also be introduced? Should be an easy enough change. Thank you for comments / suggestions / ideas!

2 1

eqType modulo associated types?
by Conal Elliott 16 Sep '19

16 Sep '19

It looks to me like `eqType` accounts for type synonyms but not associated types. Is there a variant that compares modulo associated types, or perhaps a type normalizing operation to apply before `eqType`? Thanks, - Conal

3 3

GHC/StgToCmm/Con.hs
by Shayne Fletcher 14 Sep '19

14 Sep '19

Anyone understand what's going on here (HEAD) on Windows? ``` 2019-09-14T19:06:40.1438283Z # Running: git clone https://gitlab.haskell.org/ghc/ghc.git 2019-09-14T19:06:40.1879263Z Cloning into 'ghc'... 2019-09-14T19:07:02.4403717Z error: unable to stat just-written file compiler/GHC/StgToCmm/Con.hs: No such file or directory ``` -- *Shayne Fletcher* Language Engineer */* +1 917 699 7663 *Digital Asset* <https://digitalasset.com/>, creators of *DAML <https://daml.com/>* -- This message, and any attachments, is for the intended recipient(s) only, may contain information that is privileged, confidential and/or proprietary and subject to important terms and conditions available at http://www.digitalasset.com/emaildisclaimer.html <http://www.digitalasset.com/emaildisclaimer.html>. If you are not the intended recipient, please delete this message.

3 4

Can someone help my test my MR on *BSD?
by John Ericson 12 Sep '19

12 Sep '19

MR !1699 deletes some code in the compiler about passing -l flags for the RTS. This is a layer violation and hopefully a useless one; the RTS config file in the package db lists external library dependencies just like any other package so the -l flags should be included because that alone. CI passes, so everything works on Linux, Darwin, and Windows. Ben Gamari wanted to double check the behavior on the *BSDs, since the HAVE_* configure results might reasonably vary there. I got a FreeBSD VM and started installing the requisite ports, but I've never used any BSD before and don't have much time to learn it---say how to stop to the TUI pop-ups that prevent me building everything overnight. If someone with an up-and-running development environment could quickly give this change a spin, that would be much appreciated. Fingers crossed, and thanks in advanced, John P.S. If said BSD angel wants to contribute a machine for the official CI, so neither me or anyone else is tempted to bug them again, that would also be great :) !1699: https://gitlab.haskell.org/ghc/ghc/merge_requests/1699/diffs

1 0

Using head.hackage
by Alejandro Serrano Mena 11 Sep '19

11 Sep '19

Dear GHC devs, As part of our work in "quick look impredicativity", we would like to test whether any package breaks. I've read about "head.hackage" as a way to test this fact, but I could not find any information. Could somebody point me in the right direction? If it does not exist yet, I'll try to summarize whatever I learn in the wiki for future reference. Thanks in advance, Alejandro

4 7

Min closure payload size?
by Ömer Sinan Ağacan 11 Sep '19

11 Sep '19

Hi, I was trying to understand why some info tables that have no ptrs and nptrs like GCD_CAF end up with 1 nptrs in the generated info table and found this code in Constants.h: /* ----------------------------------------------------------------------------- Minimum closure sizes This is the minimum number of words in the payload of a heap-allocated closure, so that the closure has enough room to be overwritten with a forwarding pointer during garbage collection. -------------------------------------------------------------------------- */ #define MIN_PAYLOAD_SIZE 1 We use this in a few places in the compiler and add at least one word space in the payload. However the comment is actually wrong, forwarding pointers are made by tagging the info ptr field so we don't need a word in the payload for forwarding pointers. I tried updating this as 0 but that caused a lot of test failures (mostly in GHCi). I'm wondering if I'm missing anything or is it just some code assuming min payload size 1 without using this macro. Any ideas? Ömer

3 6

Back from holiday
by Ben Gamari 10 Sep '19

10 Sep '19

Hi everyone, As I've mentioned to many of you, I spent the last few weeks at ICFP, Munihac, and bikepacking in the Alps. However, I am now back in the States and will be moving back to my usual work schedule in the coming days. Needless to say, I will make a best effort to catch up on email but if there is any particular request that you fear I may have overlooked please do send it again. Cheers, - Ben

1 0

Contributor experience report
by Bryan Richter 08 Sep '19

08 Sep '19

Hi all, Today I picked up an old worktree that hadn't seen much action for a couple months and tried to get the developer environment running again. I believe I ran into issues in Cabal, Hadrian, and/or ghc.nix that may be interesting to some. I wrote up the event log here: https://gitlab.haskell.org/snippets/1475 It all ends well; I am now successfully running hadrian/ghci.sh via ghcid within the ghc.nix shell once again. :) -Bryan P.S.: I believe Shayne recommended using the devel2 build flavor for most kinds of work. At Zurihac I ended up rebuilding the whole tree a couple times because that flavor isn't the default (once for Make, once for Hadrian). I will not in any way suggest changing the defaults, but perhaps some script or other could prompt the user to make their choice, if they haven't already?

1 0

IntRep etc
by Simon Peyton Jones 07 Sep '19

07 Sep '19

Omer, Ben There's a little cluster of tickets that it'd be really good to nail. * !1381/#16893: inline unsafeCoerce# in CorePrep * #17041: coercion safety in STG Lint * !1552: tidy up linting on unsafe coercions * #16964: definition of Int# * #17026: definition of IntRep They are all somehow tied up together. Can we kill them off together, soon? I have left comments... Simon

3 4