June 2018 - ghc-devs - Haskell.org

Literature Review of GHC Core
by Ningning Xie 26 Jun '18

26 Jun '18

Hi all, I was recently doing some GHC Core related stuff, and I found that it might be useful to have a theoretical understanding of the type system for writing better practical code. Therefore I started to read papers about GHC Core, or System FC, following the publication timeline, to try to understand each design choice involved in the type system. I wrote a simple literature review that I think might be useful for people like me. If you're thinking about learning System FC, you might find it helpful: https://github.com/xnning/GHC-Core-Literature-Review/blob/master/doc/doc.pdf Feedbacks are welcome. Best, Ningning

2 1

RE: Strace
by Simon Peyton Jones 26 Jun '18

26 Jun '18

Great. https://ghc.haskell.org/trac/ghc/ticket/15313#ticket is created. I don’t know how to force them to be “cpu multirace on windows”. If you could do that sometime it’d be great. No rush. Thank you! Simon From: Phyx <lonetiger(a)gmail.com> Sent: 26 June 2018 06:01 To: Simon Peyton Jones <simonpj(a)microsoft.com> Subject: Re: Strace Hi Simon, Thanks for the log, that does give a clue. The command fails with setup.exe: 'C:/code/HEAD/inplace/bin/ghc-pkg.exe' exited with an error: ... rule-defining-plugin-0.1: cannot find any of ["libHSrule-defining-plugin-0.1-GxqqrdsQ5NRK9hAhEkvz8Z.a","libHSrule-defining-plugin-0.1-GxqqrdsQ5NRK9hAhEkvz8Z.p_a","libHSrule-defining-plugin-0.1-GxqqrdsQ5NRK9hAhEkvz8Z-ghc8.5.20180616.so<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2FlibHSrule-d…>","libHSrule-defining-plugin-0.1-GxqqrdsQ5NRK9hAhEkvz8Z-ghc8.5.20180616.dylib","HSrule-defining-plugin-0.1-GxqqrdsQ5NRK9hAhEkvz8Z-ghc8.5.20180616.dll"] on library path (use --force to override) make[2]: *** [Makefile:18: package.T10420] Error 1 so it's the `setup install` from https://github.com/ghc/ghc/blob/c2783ccf545faabd21a234a4dfc569cd856082b9/te…<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com…> failing. unfortunately, all those tests run with -v0 which is annoying because now the verbosity of the testsuite doesn't control that of these tests. I'm not sure why these commands fail under heavy load though. I'll need to dive into the source of ghc-pkg to figure out what's happening. Notice that all the framework failures are these plugin tests which modify a package database. A wild guess is that ghc-pkg tries to take a lock on all package-databases or something when it's mutating one. But I'm not intimately familiar with the package store and this doesn't explain why it doesn't happen on Linux. for now one solution I can propose is to create a ticket to track these and mark these tests as cpu multirace on Windows, which will force them to run sequentially. I'll try to take a look at ghc-pkg this week and if I don't figure anything out I'll force the tests sequential on the short term. Cheers, Tamar On Mon, Jun 25, 2018 at 4:08 AM, Simon Peyton Jones <simonpj(a)microsoft.com<mailto:simonpj@microsoft.com>> wrote: Tamar I tried this TEST_VERBOSITY="VERBOSE=3" sh validate --fast --no-clean >& log in the root directory. I get the framework failures, but I’m not sure the verbosity-control worked. Log attached SImon

2 1

GHC API question: resolving dependencies for modules
by Dmitriy Kovanikov 26 Jun '18

26 Jun '18

Hello! I’m trying to use GHC as a library. And my goal is to be able to gather information about where each function or data type came from. I’ve started by simply calling `getNamesInScope` function and observing its result. Here is my code: * Main.hs: https://lpaste.net/9026688686753841152 <https://lpaste.net/9026688686753841152> And here is the code for my test modules: * test/X.hs: https://lpaste.net/6844657232357883904 <https://lpaste.net/6844657232357883904> * test/Y.hs: https://lpaste.net/8673289058127970304 <https://lpaste.net/8673289058127970304> Unfortunately, my implementation doesn't work since I’m not very familiar with GHC API. And I see the following errors after executing my `Main.hs` file (I’m using ghc-8.2.2): * error messages: https://lpaste.net/3316737208131518464 <https://lpaste.net/3316737208131518464> Could you please point me to places or parts of GHC API or some documentation about module dependencies and how to make ghc see imports of other modules? I can’t find simple and small enough usage example of this part of the library. Thanks in advance, Dmitrii Kovanikov

4 5

Scavenging SRTs in scavenge_one
by Ben Gamari 26 Jun '18

26 Jun '18

Hi Simon, I'm a bit confused by scavenge_one; namely it doesn't scavenge SRTs. It appears that it is primarily used for remembered set entries but it's not at all clear why this means that we can safely ignore SRTs (e.g. in the FUN and THUNK cases). Can you shed some light on this? Cheers, - Ben

3 10

Can a TC plugin create new skolem vars when simplifying Givens?
by Nicolas Frisby 24 Jun '18

24 Jun '18

I'm still spending the occasional weekend working on a type checker plugin for row types (actually "set" types at first, but I haven't thought of a less ambiguous term for that). One point of complexity in the plugin has to do with creating fresh variables when simplifying Givens. Some constraints are traditionally simplified by introducing a fresh variable. For Wanted constraints, that's easy (newFlexiTyVar). For Givens, though, I haven't figured out how to do it. This email is just to ask these two questions: 1) Is there a function to add a new skolem variable when simplifying Givens? 2) Assuming not, is there a strong reason for there to never be such a function? Here's a small indicative example. In the "simplify Givens" step, the plugin receives [G] (p `Union` Singleton A) ~ (q `Union` Singleton B) and I would ideally simplify that to [G] p ~ (x `Union` Singleton B) [G] q ~ (x `Union` Singleton A) for some fresh skolem variable x. But I don't see how to properly create a fresh skolem variable in the Givens. If these were Wanteds, I would just use newFlexiTyVar. I think this is analogous to a hypothetical type checker plugin that eta expands tuples. If we were to simplify [G] ... (x :: (k1,k2)) ... to [G] ... '(x1 :: k1,x2 :: k2) ... we'd have to generate x1 and x2 somehow. The only method I'm aware of for that is to use Fst x and Snd x instead (ie type families). That might be acceptable for the tuple expansion example, but I'm very reticent to use something like that for the set types plugin. I have a plan to get by without creating these variables when simplifying Givens, but it's not simple. I'd be delighted if it were possible to create them. Hence my two questions listed above. Thank you for your time. -Nick

1 0

atomicWriteIORef
by David Feuer 22 Jun '18

22 Jun '18

Currently, atomicWriteIORef is implemented using atomicModifyIORef. That seems pretty heavy for the job, allocating closures and forcing thunks. Is there a reason not to do it with casMutVar#?

2 1

Plan for GHC 8.6.1
by Ben Gamari 22 Jun '18

22 Jun '18

Hello fellow lazy purists, With GHC 8.4.2 out the door, it is time to begin looking forward to 8.6.1. In keeping with our six-month release schedule, this release will be targetted for early-September, with the stable branch being cut in mid-to-late June. Remarkably, this is only 6 weeks away. If you have patches that you would like to see in 8.6.1, please do put them up on Phabricator and the 8.6.1 status page [1] in the coming weeks to ensure that there is sufficient time for review. If you have a patch which you are concerned won't make the cut-off, do say something. Cheers, - Ben [1] https://ghc.haskell.org/trac/ghc/wiki/Status/GHC-8.6.1

4 4

How do I make a constraint tuple in Core?
by Ryan Scott 19 Jun '18

19 Jun '18

I'm currently working on some code in which I need to produce a Core Type that mentions a constraint tuple. I thought that there must surely exist some way to construct a constraint tuple using the GHC API, but to my astonishment, I could not find anything. The closest thing I found was mk_tuple [1], which gives you the ability to make boxed and unboxed tuples, but not constraint tuples. I then thought to myself, "But wait, PartialTypeSignatures has to create constraint tuples, right? How does that part of the code work?" To my horror, I discovered that PartialTypeSignatures actually creates *boxed* tuples (see mk_ctuple here [2]), then hackily treats them as constraint tuples, as explained in Note [Extra-constraint holes in partial type signatures] [3]. I tried reading that Note, but I couldn't follow the details. Is there a simpler way to create a constraint tuple that I'm not aware of? Ryan S. ----- [1] http://git.haskell.org/ghc.git/blob/676c5754e3f9e1beeb5f01e0265ffbdc0e6f49e… [2] http://git.haskell.org/ghc.git/blob/676c5754e3f9e1beeb5f01e0265ffbdc0e6f49e… [3] http://git.haskell.org/ghc.git/blob/676c5754e3f9e1beeb5f01e0265ffbdc0e6f49e…

2 4

Use of -dead_strip_dylibs by default makes the -framework flag appear broken.
by Travis Whitaker 18 Jun '18

18 Jun '18

Hello Haskell Friends, GHC always passes <https://github.com/ghc/ghc/blob/master/compiler/main/DriverPipeline.hs#L1888> -dead_strip_dylibs to the linker on macOS. This means that Haskell programs that use Objective-C-style dynamic binding (via objc_getClass or similar) won't actually be able to find the Objective-C methods they need at runtime. Here's an example illustrating the problem: Consider this small example program: #include <stdio.h> extern void *objc_getClass(char *n); void test_get_class(char *n) { void *cp = objc_getClass(n); if(cp == NULL) { printf("Didn't find class %s\n", n); } else { printf("Found class %s\n", n); } } int main(int argc, char *argv[]) { test_get_class(argv[1]); return 0; } Building like this: clang -o hasclass main.c -lobjc -L/usr/lib -framework Foundation -F /System/Library/Frameworks/ Yields an executable that works like this: $ ./hasclass NSObject Found class NSObject $ ./hasclass NSString Found class NSString $ ./hasclass NSDate Found class NSDate otool shows that we're linked against Foundation properly: hasclass: /usr/lib/libobjc.A.dylib (compatibility version 1.0.0, current version 228.0.0) /System/Library/Frameworks/Foundation.framework/Versions/C/Foundation (compatibility version 300.0.0, current version 1452.23.0) /usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 1252.50.4) Now consider this equivalent Haskell example: module Main where import Foreign.C.String import Foreign.Ptr import System.Environment foreign import ccall objc_getClass :: CString -> IO (Ptr a) testGetClass :: String -> IO () testGetClass n = withCString n $ \cn -> do cp <- objc_getClass cn let m | cp == nullPtr = "Didn't find class " ++ n | otherwise = "Found class " ++ n putStrLn m main :: IO () main = getArgs >>= (testGetClass . head) Building like this: ghc -o hasclass Main.hs -lobjc -L/usr/lib -framework foundation -framework-path /System/Library/Frameworks/ Yields an executable that works like this: $ ./hasclass NSObject Found class NSObject $ ./hasclass NSString Didn't find class NSString $ ./hasclass NSDate Didn't find class NSDate otool shows that our load commands for Foundation are missing: hasclass: /usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 1252.50.4) /usr/lib/libobjc.A.dylib (compatibility version 1.0.0, current version 228.0.0) /nix/store/7jdxjpy1p5ynl9qrr3ymx01973a1abf6-gmp-6.1.2/lib/libgmp.10.dylib (compatibility version 14.0.0, current version 14.2.0) /usr/lib/libiconv.2.dylib (compatibility version 7.0.0, current version 7.0.0) Interestingly, the testGetClass function will work just fine in GHCi, since it always loads all of the shared objects and frameworks it's asked to. As far as I can tell the only way to get a hasclass executable with the correct behavior is to do the final linking manually with Clang. My understanding is that this behavior was introduced to work around symbol count limitations introduced in macOS Sierra. It would be nice to wrap the frameworks passed to the linker in some flags that spares them from -dead_strip_dylibs. I haven't found such a feature in my limited digging around, but perhaps someone who knows more about systems programming on macOS will have an idea. Statically linking against the system frameworks would be a workable stopgap solution, but I have yet to find an easy way to do that. I'm curious what others' thoughts are on this issue; it's very difficult to call Objective-C methods from Haskell (without generating Objective-C) without a fix for this. Regards, Travis Whitaker

3 2

Strace
by Simon Peyton Jones 18 Jun '18

18 Jun '18

Tamar I'm getting megabytes of output from 'sh validate' on windows. It looks like this 629 151745 [main] sh 2880 fhandler_base::fhaccess: returning 0 291 152036 [main] sh 2880 faccessat: returning 0 7757 159793 [main] sh 2880 fhandler_base_overlapped::wait_overlapped: wfres 0, wores 1, bytes 7 179457 1608947 [main] make 11484 fhandler_base_overlapped::wait_overlapped: wfres 0, wores 1, bytes 7 99 159892 [main] sh 2880 fhandler_base_overlapped::wait_overlapped: normal write, 7 bytes ispipe() 1 180 1609127 [main] make 11484 fhandler_base_overlapped::wait_overlapped: normal read, 7 bytes ispipe() 1 139 160031 [main] sh 2880 write: 7 = write(1, 0x6000396A0, 7) 142 1609269 [main] make 11484 fhandler_base::read: returning 7, binary mode 139 1609408 [main] make 11484 read: 7 = read(5, 0x60005B4B0, 7) 136 1609544 [main] make 11484 read: read(5, 0x60005B4B7, 193) blocking 4693 164724 [main] sh 2880 set_signal_mask: setmask 0, newmask 80000, mask_bits 0 but with hundreds of thousands of lines. (I have not counted) I believe that it may be the result of this line, earlier in the log cd "/c/Users/simonpj/AppData/Local/Temp/ghctest-8fa9s6rk/test spaces/./plugins/plugins07.run" && strace $MAKE -s --no-print-directory -C rule-defining-plugin package.plugins07 TOP=/c/code/HEAD/testsuite# Note the strace. That in turn was added in your commit commit 60fb2b2160aa16194b74262f4df8fad5af171b0f Author: Tamar Christina <tamar(a)zhox.com> Date: Mon May 28 19:34:11 2018 +0100 Clean up Windows testsuite failures Summary: Another round and attempt at getting these down to 0. Could you perhaps have made a mistake here? Currently validate is unusable. Thanks! Simon

2 6