July 2002 - Glasgow-haskell-users

Deadlock detection different in ghc-5.04?
by Volker Stolz 22 Jul '02

22 Jul '02

Hi, for some concurrency abstractions I decided to use deadlock detection by means of exceptions which didn't work out as expected. Below is some simplifed source code showing the problem: My assumption was since GHC should be able to detect that the child still has a reference to mv_should_work -- although in an exception handler -- the RTS would never decide to kill the other thread. Unluckily, this assumption turned out to be false. Did I overestimate GHC's capabilities or is this a bug? Especially as ghc-5.02.2 shows the desired behaviour (you just need to drop the "Control.")... Instead of exiting silently, the program terminates in the exception handler in the main thread with ghc-5.04. \begin{code} module Main where import Concurrent import qualified Control.Exception main = do mv_should_work <- newEmptyMVar mv_deadlock <- newEmptyMVar forkIO $ do Control.Exception.catch (takeMVar mv_deadlock) -- deadlock (\ e -> putMVar mv_should_work ()) yield >> yield >> yield Control.Exception.catch (takeMVar mv_should_work) (\ e -> putStrLn $ "main caught: " ++ (show e)) \end{code} -- Volker Stolz * http://www-i2.informatik.rwth-aachen.de/stolz/ * PGP * S/MIME http://news.bbc.co.uk: `Israeli forces [...], declaring curfews that confine more than 700,000 people to their homes.'

1 0

FreeBSD port for 5.04 now committed
by Simon Marlow 22 Jul '02

22 Jul '02

The FreeBSD folks have been kind enough to commit the update for the ghc port to 5.04. FreeBSD users: cvsup your /usr/ports, cd /usr/ports/lang/ghc && make install, or wait for the package to appear on your local mirror site and try 'pkg_add -r ghc'. Cheers, Simon

1 0

RE: fno-implicit-prelude and literal numeric patterns
by Simon Peyton-Jones 22 Jul '02

22 Jul '02

| The user's guide is silent about which version of Eq is used | for literal patterns, but I assume that it follows the (n+k) | example and so Prelude Eq is used for the overloaded use of ==. That's right. I'll add a note to that effect in the users guide. | What is the reason for using Prelude.Ord (and Prelude.Eq)? | | This seems very limiting since you can replace Num but you | can't replace Eq, and moreover, your new versions of the | Numeric classes must be subclasses of Prelude.Eq, rather than | another Eq. |... | Is there are strong reason for avoiding the alternative: | "whatever == and >= are in scope"? Perhaps it is the | if-then-else that must refer to Prelude.Bool? Well, I had to stop somewhere. (As I have previously remarked on this thread, it is hard to make *everything* rebindable.) With more effort one could make more things rebindable. My goal was to make numerics completely rebindable; the current omission is (only) the handling of defaults. At the moment I simpy don't know what a good 'final' design might be, and I'm pretty reluctant to develop this feature incrementally. If a consensus emerges, then yes (unless it's a heart-and-lung job) I'll implement it or help one of you to do so. Simon

1 0

RE: Docs missing?
by Simon Marlow 22 Jul '02

22 Jul '02

The RPMs of documentation are currently not quite right - they don't contain any of the Haddock-generated docs or the index.html that ties it all together. We know about the problem, but no-one has had the time (or in my case, the RPM-savvyness) to fix it yet. Michael Weber said: > In the ghc5-doc Debian packages these files live under: > /usr/share/doc/ghc5-doc/html/users_guide/ > > To install the html docs I used: > > $(MAKE) -C build-stage$(MAXSTAGE) \ > install-docs \ > prefix=$(CURDIR)/$(destdir)/$(prefix) \ > datadir=$(CURDIR)/$(destdir)/$(prefix)/share/doc/$(package) > > with 'SGMLDocWays="html dvi ps"' in mk/build.mk. Yes, that's the right way to do it. The docs are now set up so that when installed properly you should get an index.html in $(datadir)/ghc-5.04/html which points to the rest of the HTML docs. > Unfortunately, the link does not work here either, since the file > users-guide.html is named book1.html :( . But that's another > problem, I think, and likely related to docbook-utils. Hmm, that's rather unfortunate. I think we might have to include some more conditional bits in index.html.in so that the links point to the right place in the case when the DocBook output ends up in book1.html instead of users_guide.html. > Unless anyone objects, I'll delay fixed Debian packages until > 5.04.1 is released. No problem. Cheers, Simon

1 0

RE: ANNOUNCE: GHC version 5.04 released
by Simon Marlow 22 Jul '02

22 Jul '02

> I've patched up an install script for OpenBSD users of GHC who: > > - would like to install it somewhere other than /usr/local > - do not have root priviledges on their machine > > Note that uninstallation must be by hand if you use this script, as we > bypass the pkg system. Could this be done using GHC's normal binary distribution mechanism? In a build tree you do the following: - add the line 'BIN_DIST=1' to mk/build.mk - build everything (inc. profiling libraries if required) - say 'make binary-dist Project=Ghc' at the top level - tar up the contents of ghc-5.04 and ship it this gives you a tarball which can be unpacked and used in-place or installed elsewhere in the filesystem. Cheers, Simon

1 0

Re: DeepSeq
by Phil Trinder 22 Jul '02

22 Jul '02

Dean, George, >Would it be possible to bring the DeepSeq library into the libraries >distributed with GHC? (I think Dean Herington is responsible for it.) > >Of course it's easy enough to drop it into one's own program (I am just >about to do this) but >(1) It is fairly common to want to force deeper evaluation. >(2) DeepSeq is simple enough to be dropped in the GHC distribution, without >it causing much trouble or making it much bigger. >(3) At the same time, it is not so simple that it can be reimplemented in a >couple of lines. We use 'Evaluation Strategies' to control evaluation in parallel and distributed Haskells. The 'rnf' (reduce to normal form) strategy is equivalent to DeepSeq, and the Strategies library is already available in GHC (-syslib concurrent). More details in Trinder P.W. Hammond K. Loidl H-W. Peyton Jones S.L. Algorithm + Strategy = Parallelism. Journal of Functional Programming, 8(1) (January 1998) pp~23-60. Phil -------------------------------------------------- Phil Trinder Department of Computing and Electrical Engineering Heriot Watt University Riccarton Edinburgh, EH14 4AS E-mail: trinder(a)cee.hw.ac.uk Teleph: +44 (0)131 451 3435 Depart: +44 (0)131 451 3328 Fasmly: +44 (0)131 451 3327 Intrnt: http://www.cee.hw.ac.uk/~trinder

1 0

Docs missing?
by Jorge Adriano 21 Jul '02

21 Jul '02

Hi, I installed Ghc 5.04 (SuSE rpms), and was checking the documentation. I opened what seemed to me the main index file: file:/usr/share/doc/packages/ghc/docs/index.html There we have links to: 1. The User's Guide: file:/usr/share/doc/packages/ghc/docs/users_guide/users-guide.html The correct link is actually: file:/usr/share/doc/packages/ghc/docs/users_guide/users_guide/users-guide.html In file:/usr/share/doc/packages/ghc/docs/users_guide/ there are the what appears to be the "users-guide" sources. I don't know wether they were included by mistake or not. 2. Libraries, namely Base: file:/usr/share/doc/packages/ghc/docs/base/index.html Haskell98: file:/usr/share/doc/packages/ghc/docs/haskell98/index.html Network: file:/usr/share/doc/packages/ghc/docs/network/index.html Haskell-src: file:/usr/share/doc/packages/ghc/docs/haskell-src/index.html None of this links work. I can't find the libraries documentation anywhere in file:/usr/share/doc/packages/ghc/docs/ Same goes to hslibs - the "old haskell libraries". file:/usr/share/doc/packages/ghc/docs/hslibs/book-hslibs.html I listed the files in the RedHat ghc documentation package, and this seemed to be missing too. What am I missing here? Just the Ghc docs or something else? :) J.A.

2 1

Re: ANNOUNCE: GHC version 5.04 released
by Ashley Yakeley 20 Jul '02

20 Jul '02

At 2002-07-19 20:25, Michael Weber wrote: >The archive can be used by APT after adding the following two lines to >/etc/apt/sources.list: >----------------------------------------------------------------------------- >deb http://lambda.foldr.org/~michaelw/debian-unoff unstable . >deb-src http://lambda.foldr.org/~michaelw/debian-unoff unstable . >----------------------------------------------------------------------------- >(Don't forget to add the dot at the end of the line as well!) I got this: Err http://lambda.foldr.org unstable/. ghc5-doc 5.04-1 404 Not Found Failed to fetch http://lambda.foldr.org/~michaelw/debian-unoff/dists/unstable/./binary-i386 /ghc5-doc_5.04-1_all.deb 404 Not Found E: Unable to fetch some archives, maybe run apt-get update or try with --fix-missing? -- Ashley Yakeley, Seattle WA

2 1

re: fno-implicit-prelude and literal numeric patterns
by Bernard James POPE 20 Jul '02

20 Jul '02

Hi all, I wrote: > I know I'm becoming a pest with this, but another question about > -fno-implicit-prelude: > > What happens with literal numeric patterns? Ok the various bits and pieces that I think are relevent: The Haskell Language Report: 3.17.3 Formal Semantics of Pattern Matching (h) case v of { k -> e; _ -> e' } = if (v==k) then e else e' where k is a numeric, character, or string literal. (s) case v of { x+k -> e; _ -> e' } = if v >= k then (\x -> e) (v-k) else e' where k is a numeric literal Rule (h) in Figure 4 involves the overloaded operator ==; it is this rule that d efines the meaning of pattern matching against overloaded constants. The Ghc users guide: 7.5.4. Rebindable syntax In an n+k pattern, the standard Prelude Ord class is still used for comparison, but the necessary subtraction uses whatever "(-)" is in scope (not "Prelude.(-)"). The user's guide is silent about which version of Eq is used for literal patterns, but I assume that it follows the (n+k) example and so Prelude Eq is used for the overloaded use of ==. What is the reason for using Prelude.Ord (and Prelude.Eq)? This seems very limiting since you can replace Num but you can't replace Eq, and moreover, your new versions of the Numeric classes must be subclasses of Prelude.Eq, rather than another Eq. Is there are strong reason for avoiding the alternative: "whatever == and >= are in scope"? Perhaps it is the if-then-else that must refer to Prelude.Bool? Cheers, Bernie.

1 0

Re: ANNOUNCE: GHC version 5.04 released
by Ashley Yakeley 19 Jul '02

19 Jul '02

At 2002-07-11 11:22, Eray Ozkural wrote: >> A Windows installer for 5.04 is now available via the GHC downloads >> page: http://haskell.org/ghc/download_ghc_504.html > >/me pings debian people. we can't lag behind windows, come on ;) I'm sure Michael Weber is working on it even as we speak... -- Ashley Yakeley, Seattle WA

2 2