April 2016 - ghc-devs

Reification of out-of-scope variables?
by Boespflug, Mathieu 19 Apr '16

19 Apr '16

Hi all, the community is increasingly using quasiquotation as a means to interoperate with foreign languages. E.g. language-c-inline, inline-c and inline-r allow calling foreign functions using foreign language syntax, rather than "foreign import" declarations. There are efforts underway for Java, Javascript and other languages too. The pattern common to most of these libraries is: * collect the content of all quasiquotes that appear in a module, generate a compilable foreign source file based on that, and then link the object code resulting from compiling that. When the foreign language is itself statically typed, we need to make sure we generate code with proper type annotations, including for any antiquotation variables. In older versions of GHC, we could derive the type annotations from the inferred Haskell types, but this is not possible since 7.8 because the types of any variable in the current "declaration group" are not made available. So libraries like inline-c ask for some extra verbosity from the user: mycossin1 :: Double -> IO Double mycossin1 somex = [cexp| double { cos($(double somex)) * sin($(double somex)) } |] The type annotations inside the quasiquote are redundant with the explicitly provided type signature. C types are typically short enough, but other languages (e.g. C++, Java), have really long fully qualified type names, so the extra annotations are a cost. The are good reasons why types aren't available from within a declaration group (it was possible to observe the order in which type inference works), hence why the extra annotations are necessary. But by the time type checking of the whole module is finished, types are fixed once and for all. So the question is: * Could we make it possible to query the types of local variables at the end of type checking? The reason I ask is: with 'addModFinalizer' and friends that TH provides, we're tantalizingly close to being able to generate foreign code wrappers that have the right types that match those of the antiquotation variables present in a module. Template Haskell already provides 'addModFinalizer'. You feed it a Q action, which will run once the whole module is type checked. If at that point we could ask, "what's the type of somex (above)?", then we could let the user write mycossin2 :: Double -> IO Double mycossin2 somex = [cexp| cos($somex) * sin($somex) |] and yet still generate a C file capturing the content of above quasiquote with the right types: double module_foo_qq1(double v1) { return (cos(v1) * sin(v1)); } since we know that somex :: Double and that the Haskell type "Double" corresponds to C's "double" primitive type. Bound variables have a unique name associated to them, which we can get hold of in Template Haskell using the 'var syntax, but f x = $(let name = 'x in addModFinalizer (reify name >>= \info -> runIO (print info)) >> [| x |]) results in a compiler error, The exact Name `x' is not in scope Probable cause: you used a unique Template Haskell name (NameU), perhaps via newName, but did not bind it If that's it, then -ddump-splices might be useful because by the time the TH finalizer runs, we're no longer in the scope of x. But since the names of the local variables are unique, could it possibly make sense to allow calling reify on them, even outside of their scope, so as to get at their type? Best, Mathieu

6 17

RE: [GHC] #11824: GHC error in desugarer lookup
by Simon Peyton Jones 19 Apr '16

19 Apr '16

Ben I'm offline, so can't reply on Phab. Calling mkTypeableBinds in tcRnHsBootDecls isn't quite right, because it'll generate lots of *bindings* whereas all we want is the *Ids* in the GlobalValEnv. (It's possible that tcg_binds is ignored in the boot-module case, but it's unsavoury to generate the bindings at all. They are generated when we process the main module.) I suggest the easiest fix is, in mkModIdBinding and mkTypeableTyConBinds, guard the binding-generation with a test for whether isHsBootOrSig is true. Simon | -----Original Message----- | From: ghc-tickets [mailto:ghc-tickets-bounces@haskell.org] On Behalf Of GHC | Sent: 13 April 2016 12:06 | Cc: ghc-tickets(a)haskell.org | Subject: Re: [GHC] #11824: GHC error in desugarer lookup | | #11824: GHC error in desugarer lookup | -------------------------------------+------------------------------------- | Reporter: darchon | Owner: | Type: bug | Status: patch | Priority: normal | Milestone: 8.0.1 | Component: Compiler | Version: 8.0.1-rc3 | Resolution: | Keywords: | Operating System: Unknown/Multiple | Architecture: | Type of failure: Compile-time | Unknown/Multiple | crash | Test Case: | Blocked By: | Blocking: | Related Tickets: | Differential Rev(s): Phab:D2107, | Wiki Page: | Phab:D2108 | -------------------------------------+------------------------------------- | Changes (by bgamari): | | * status: new => patch | * differential: => Phab:D2107, Phab:D2108 | | | -- | Ticket URL: | <https://na01.safelinks.protection.outlook.com/?url=http%3a%2f%2fghc.haskell. | org%2ftrac%2fghc%2fticket%2f11824%23comment%3a6&data=01%7c01%7csimonpj%40064d | .mgd.microsoft.com%7cd4862d0ed83c493164fb08d3638c0642%7c72f988bf86f141af91ab2 | d7cd011db47%7c1&sdata=JaXInqbrHs1nrAM9JsMUfX6J6hfbzJHI2k5KKRbH9i0%3d> | GHC | <https://na01.safelinks.protection.outlook.com/?url=http%3a%2f%2fwww.haskell. | org%2fghc%2f&data=01%7c01%7csimonpj%40064d.mgd.microsoft.com%7cd4862d0ed83c49 | 3164fb08d3638c0642%7c72f988bf86f141af91ab2d7cd011db47%7c1&sdata=S85zedvr8ceEm | 4UeWtejgMl%2buRm4XMrgdWiHwcXDVuU%3d> | The Glasgow Haskell Compiler | _______________________________________________ | ghc-tickets mailing list | ghc-tickets(a)haskell.org | https://na01.safelinks.protection.outlook.com/?url=http%3a%2f%2fmail.haskell. | org%2fcgi-bin%2fmailman%2flistinfo%2fghc- | tickets&data=01%7c01%7csimonpj%40064d.mgd.microsoft.com%7cd4862d0ed83c493164f | b08d3638c0642%7c72f988bf86f141af91ab2d7cd011db47%7c1&sdata=WywP4RhTbL5cEfl%2b | xztoUe1EYuHPthiCb3vS2%2b%2fP4F8%3d

2 6

Re: Should TH TExp be able use the Q monad
by Carter Schonwald 18 Apr '16

18 Apr '16

This thread is reminding me of the metaprograming ux that agda has as of the new 2.5 release, where there's a type checking monad for proposed terms and the ability to choose to do error handling when a proposed term fails to type check. Note that their meta ast is unindexed, so the program fragments are constructed in a sort of untyped debrujin data model of agda terms On Apr 18, 2016 10:04 AM, "Richard Eisenberg" <eir(a)cis.upenn.edu> wrote: On Apr 18, 2016, at 9:45 AM, Simon Peyton Jones <simonpj(a)microsoft.com> wrote: > Well, it opens up the entire issue of dependence on typechecking order and reification. Other things being equal, simple is good... Of course that's true, but other things aren't equal -- losing Q decreases the usefulness of typed TH. I agree that there is some nastiness regarding reification. We could refuse to reify something from the same group. I'm not sure how hard that would be to enforce. Or if reification were a real bear, we could provide the IO monad. Just to see how this is used, I poked around in a download of all of Hackage from September 2015. Here's what I found. - 6 packages use typed TH: clash-prelude-0.9.2, llvm-general-quote-0.2.0.0, lookup-tables-0.1.1.1, network-uri-static-0.1.0.0, refined-0.1.1.0, and validated-literals-0.2.0. - None seem to use the ability to do work in the Q (or IO) monad. - Two (lookup-tables and refined) do use the fact that the TExp constructor is exported from Language.Haskell.TH.Syntax to make an untyped TH expression and unsafely force it into a typed TH expression. If we're going to close back doors, this seems like a much wider one than access to Q. So I guess this suggests that taking typed TH out of the Q monad wouldn't be too disruptive. But saying "no reification of anything, anywhere" seems like a big sledgehammer to stop people from reifying local things whose typed haven't settled yet. Richard _______________________________________________ ghc-devs mailing list ghc-devs(a)haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs

2 2

directory and GHC 8.0.1
by Ben Gamari 18 Apr '16

18 Apr '16

Hi Phil, I noticed that directory has a master-1.2.6 branch with at least one useful looking bug-fix on it. Is this the branch that you would prefer to be included in 8.0.1? Cheers, - Ben

4 10

Whither default build variables for make files?
by Carter Schonwald 18 Apr '16

18 Apr '16

Hey all, I just did a checkout of the tip of ghc 8.0 branch and it looks like the build.mk files have seen a bit of reorganization! Now for the perf aka release build the various doc variavles seem to be implicitly set somewhere else and I got a bit lost when trying to dig them up. On the flip side : is the shake build system alternative mature enough for it to be reasonable to attempt using it on my Mac ? Many thanks

3 3

Should TH TExp be able use the Q monad
by Simon Peyton Jones 18 Apr '16

18 Apr '16

| > Hang on! The design for typed splices, describe here, | > https://ghc.haskell.org/trac/ghc/wiki/TemplateHaskell/BlogPostChanges | > says "Unlike TH, the only way to construct a value of type TExp is | with a quote. You cannot drop into do-notation, nor use explicit | construction of the values in the Exp algebraic data type. That | restriction limits expressiveness, but it enables the strong typing | guarantees." | > | > So why does the above work? $$(e) requires a TExp, and do-notation | doesn’t produce a TExp. | | Indeed this is true -- this is what that page says. But it's not what's | implemented: when I say $$( _ ) :: Bool, GHC tells me that the hole has | type Q (TExp Bool). | There still is no way to create a TExp other than to use a type TH quote Humph. I suppose that provided you can't forge a TExp, there's no way to splice in a type-incorrect program; and having Q available lets you (say) do input/output or consult the context to decide which of two quotes to return. We could do this selectively. For example we'd certainly need (TExp t) to be able to fail > bool :: String -> TExp Bool > bool "true" = [|| True ||] > bool "false" = [|| False ||] > bool _ = failTexp "not a bool" My instinct is to make it less expressive, though, and only allow (TExp t) as the argument of $$. Does anyone care either way? I suppose we'd better open a ticket for this. Simon | | quote. | | Addressing your other message: a typed quasiquoter would be somewhat | limited, but not utterly silly. For example, this works: | | > bool :: String -> Q (TExp Bool) | > bool "true" = [|| True ||] | > bool "false" = [|| False ||] | > bool _ = fail "not a bool" | > | > -- and then, in another module because of the stage restriction: | > yes :: Bool | > yes = $$(bool "true") | | Now `bool` could be a typed quasiquoter. | | I don't know whether any of this is worth implementing, but it's not, a | priori, a terrible idea. | | Richard | | > | > | * Should we consider it a bug (and file a ticket) that reification | > | in typed splices is able to observe the order of type checking, | > | just like reify used to do in untyped splices? | > | > Yes I think so!!! | > | > Simon | >

2 3

RE: Whither default build variables for make files?
by Andrey Mokhov 18 Apr '16

18 Apr '16

Hi Carter, > On the flip side : is the shake build system alternative mature enough > for it to be reasonable to attempt using it on my Mac ? Here are the limitations of the Shake build system: https://github.com/snowleopard/shaking-up-ghc#current-limitations If you are not put off by them, please give the new build system a try! Cheers, Andrey

1 0

Phabricator build failures
by Joachim Breitner 18 Apr '16

18 Apr '16

Hi, some recent commits on GHC HEAD fail to validate on phabricator, all with a segfault in the tests, but in an different tests each (or sometimes no failure at all). This this known? Greetings, Joachim -- Joachim “nomeata” Breitner mail(a)joachim-breitner.de • https://www.joachim-breitner.de/ XMPP: nomeata(a)joachim-breitner.de • OpenPGP-Key: 0xF0FBF51F Debian Developer: nomeata(a)debian.org

2 1

Spam filtering
by Ben Gamari 17 Apr '16

17 Apr '16

Hello GHCers, As you may have noticed, the GHC Trac instance is currently being hit with a significant amount of spam traffic recently. I believe I've cleared up most of the garbage and the filter configuration has been tightened so that future spam attacks will be easier to quell. Of course, this tightening means that there is a greater chance of false positives. If you do find that your ticket comments or Wiki edits are being rejected by the spam filter, please notify either Austin or me so that we can train the filter on your example. Cheers, - Ben

3 5

Re: GHC 8.0.1-rc3 source tarball availability
by Ben Gamari 15 Apr '16

15 Apr '16

Karel Gardas <karel.gardas(a)centrum.cz> writes: > Hi Ben, > > just got back from business trip and attempting to build distros for you > I've found that: > > karel@silence:~/Downloads/GHC/8.0.1/rc3$ gpg2 --verify > ghc-8.0.0.20160411-src.tar.xz.sig > Warning: using insecure memory! > gpg: Signature made April 11, 2016 03:23:45 AM CEST using RSA key ID > 97DB64AD > gpg: BAD signature from "Benjamin Gamari <ben(a)well-typed.com>" > It seems that the issue was once again my inability to convince the content delivery network to fully purge its cache. Thanks for realizing this issue. On the bright side, I think I have finally worked out how to reliably flush the cache so hopefully there will be no further issues of this tiresome sort. Unfortunately it's quite unclear what to do about your binary distribution as it's likely derived from the wrong source tree. As we will likely have to do a fourth release candidate anyways (due to the template haskell issue brought up by Iavor today), perhaps it would be best just to wait until then before sinking any more of your time into further builds. It likely won't be long before we have another source tree available although I'll leave the decision to you. Cheers, - Ben

1 0