From dmellis at gmail.com Mon May 1 17:54:42 2006 From: dmellis at gmail.com (David A. Mellis) Date: Tue May 2 03:35:24 2006 Subject: [Hat] hat: missing separator in generated config. Message-ID: I don't think this is a problem with hat (since the same thing happened when I built hmake), but since it breaks the hat build (at least on my machine), I thought I should report it. My generated config file (lib/powerpc-Darwin8/config) contains the three lines (9 through 11): GHCSYM= 604 and so when running "make", I get the error: /Users/dmellis/Desktop/hat-2.04/lib/powerpc-Darwin8/config:11: *** missing separator. Stop. This is on Mac OS X 10.4 (Tiger), with GNU Make 3.80. The file "targets/powerpc-Darwin8/ghcsym" has two blanks lines at the beginning, followed by a line with only "604" - perhaps this is the problem? From Malcolm.Wallace at cs.york.ac.uk Tue May 2 09:20:20 2006 From: Malcolm.Wallace at cs.york.ac.uk (Malcolm Wallace) Date: Tue May 2 09:16:58 2006 Subject: [Hat] hat: missing separator in generated config. In-Reply-To: References: Message-ID: <20060502142020.4b34dad7.Malcolm.Wallace@cs.york.ac.uk> "David A. Mellis" wrote: > My generated config file (lib/powerpc-Darwin8/config) > contains the three lines (9 through 11): > > GHCSYM= > > 604 Yes, this is a fault in the detection logic for the compiler version, again due to gcc-4. I assume you worked out that you just need to make it into a single line GHCSYM=604 Meanwhile, the CVS version of the configure script has been updated (some time ago) to fix the issue. I suppose it might be useful to release an updated package for Hat, just to avoid the annoyance for other people on the Apple platform. Regards, Malcolm From tatd2 at kent.ac.uk Tue May 2 11:44:14 2006 From: tatd2 at kent.ac.uk (Thomas Davie) Date: Tue May 2 11:44:17 2006 Subject: [Hat] Hat-trans or hat-lib bug Message-ID: <27121B65-015D-4CD8-A6E7-0D15B311EC6D@kent.ac.uk> Odd bug here... The file lambda.hs contains: main = print ((\x -> x * x) (2.0 :: Float)) This is the output for compiling for hat, and not for hat: Looking at the traces, the recorded trace computes the correct answer (in fact, 16 can't be found anywhere in the trace), but the incorrect answer is output. alduruch:~/Documents/Work/Hat test cases/Lambda tatd2$ rm lambda.o alduruch:~/Documents/Work/Hat test cases/Lambda tatd2$ rm Hat/lambda.o alduruch:~/Documents/Work/Hat test cases/Lambda tatd2$ rm lambda alduruch:~/Documents/Work/Hat test cases/Lambda tatd2$ hmake lambda ghc -c -o lambda.o lambda.hs ghc -o lambda lambda.o alduruch:~/Documents/Work/Hat test cases/Lambda tatd2$ ./lambda 4.0 alduruch:~/Documents/Work/Hat test cases/Lambda tatd2$ rm lambda alduruch:~/Documents/Work/Hat test cases/Lambda tatd2$ hmake -hat lambda hat-trans lambda.hs Wrote Hat/lambda.hs ghc -c -package hat -o Hat/lambda.o Hat/lambda.hs ghc -package hat -o lambda Hat/lambda.o alduruch:~/Documents/Work/Hat test cases/Lambda tatd2$ ./lambda 16.0 From tatd2 at kent.ac.uk Sat May 6 10:46:35 2006 From: tatd2 at kent.ac.uk (Thomas Davie) Date: Sat May 6 10:40:31 2006 Subject: [Hat] Is the CVS repo down? Message-ID: I've been getting "could not contact haskell.galois.com" errors when trying to do a cvs update... Is this a my connection issue, or a their connection issue? Bob From Malcolm.Wallace at cs.york.ac.uk Sat May 6 11:09:05 2006 From: Malcolm.Wallace at cs.york.ac.uk (Malcolm Wallace) Date: Sat May 6 11:02:37 2006 Subject: [Hat] Is the CVS repo down? In-Reply-To: References: Message-ID: <20060506160905.11491100.Malcolm.Wallace@cs.york.ac.uk> Thomas Davie writes: > I've been getting "could not contact haskell.galois.com" errors when > trying to do a cvs update... Is this a my connection issue, or a > their connection issue? It works for me... Regards, Malcolm From mrchebas at gmail.com Tue May 9 04:08:29 2006 From: mrchebas at gmail.com (Alexey Rodriguez Yakushev) Date: Mon May 15 05:12:03 2006 Subject: [Hat] Tracing a program that uses Control.Monad.ST Message-ID: <0004D022-DF33-11DA-A519-000393AB3466@gmail.com> Hi all, I am attempting to trace a program (Djinn) that uses library modules that are not supported by Hat, such as Control.Monad.ST . I was planning to add them to either add them to Hat before building or directly to the program itself. However, I fear that this and other modules were left out because there was some fundamental limitation of Hat when dealing with these. I don't want to discover this after hours of trying so I decided to ask here :). Cheers, Alexey P.S. Please cc me, I am not subscribed. From Malcolm.Wallace at cs.york.ac.uk Mon May 15 06:16:32 2006 From: Malcolm.Wallace at cs.york.ac.uk (Malcolm Wallace) Date: Mon May 15 06:11:01 2006 Subject: [Hat] Tracing a program that uses Control.Monad.ST In-Reply-To: <0004D022-DF33-11DA-A519-000393AB3466@gmail.com> References: <0004D022-DF33-11DA-A519-000393AB3466@gmail.com> Message-ID: <20060515111632.50c06c65.Malcolm.Wallace@cs.york.ac.uk> Alexey Rodriguez Yakushev wrote: > Hi all, I am attempting to trace a program (Djinn) that uses library > modules that are not supported by Hat, such as Control.Monad.ST . > > I was planning to add them to either add them to Hat before building > or directly to the program itself. However, I fear that this and > other modules were left out because there was some fundamental > limitation of Hat when dealing with these. I don't want to discover > this after hours of trying so I decided to ask here :). No, there is no fundamental reason why Control.Monad.ST is not currently traced. For STRefs, just follow the scheme that is already used for tracing IORefs. The rank-2 type of runST might pose a rather interesting question - I doubt whether hat-trans can generate the correct transformed type for it. However, you can probably write its transformed type by hand, with some careful thought. Regards, Malcolm