buffered output; :s -p"%s%s%s"
Hello, I noticed two things which I think might be called bugs, at least I am near doing so: The first thing really annoys me: If you input something like [0,x] where x=x you get no output at all before you interupt. I think the output is buffered, and I don't like it. This is worse in less strange examples like doing a slow calculation of a list, e.g. calculating perfect numbers by filtering all numbers. If you really want to buffer output even to stdout, please consider flushing it frequently. The other thing is less import and more like a bug, after all it lets you crash hugs quite easily. (I really guess you know it.) Just type :s -p"%s%s%s" (you may need more %s's -- three were always enough for me -- and can of course also use %i or whatever you like). The reason is obviosly the way the prompt is enabled to tell the last loded module (a feature I don't like as a default anyway), so some checking of the string given to the p-option should be done. Perhaps more complex things could be done in order to cleanly allow even more prompt tricks. I considered doing it myself, but I really can't do it this month. I don't think it matters, but I tried this with Linux and SunOS, and the hugs version calls itself 970410. Christian Sievers
Installation problems
To The Haskell-Hackers, I am trying to install ghc version 2.02 on sun4c architechture. 1 In file fptools/literate/mkdependlit I find the following lines at the top: $LIB_ARCH_DIR="/amd/church/projects/pacsoft/D/ghc-builds/1.3/opt-sparc-sun-sunos4/literate"; $LIB_DIR="/amd/church/projects/pacsoft/D/ghc-builds/1.3/opt-sparc-sun-sunos4/literate"; $TGRIND_HELPER="/usr/local/lib/tgrind/tfontedpr "; They don't seem to have been customised by `configure' or `gmake' during installation, and as our system paths are different, this seems to result in a `Giant Error' -- a rather frightening message. When I change these lines to some paths that do in fact exist, the `Giant Error' turns into two less scary errors about 2-03-notes.lit and syslib.lit not being found. I searched for these file names in the distribution but could not find them. 2 It might be worth mentioning that apart from needing the "good old 0.29 compiler" for booting the 2.03 version you also need a working unlit, and this must be correctly pointed to by the 0.29 mkdependHS1.2. I had to change the absolute path names in mkdependHS1.2 to make it succeed when calling unlit. 3 I don't know whether it is of any general significance, but during compilation of ghc/compiler/prelude/PrimOp.o the ghc-0.29 aborted with stack problems on a 6Mb stack. This I seemed to be able to fix by adding a -H16m switch in the ghc/compiler/Makefile module options list. 4 I cannot link the system! There are simply too many .o files, so when attempting ghc-0.29 -o hsc -cpp -fhaskell-1.3 -syslib ghc -fglasgow-exts -DCOMPILING_GHC -Rghc-timing -I. -IcodeGen -InativeGen -Iparser -iutils -ibasicTypes -itypes -ihsSyn -iprelude -irename -itypecheck -ideSugar -icoreSyn -ispecialise -isimplCore -istranal -istgSyn -isimplStg -icodeGen -iabsCSyn -imain -ireader -iprofiling -iparser -inativeGen -DOMIT_DEFORESTER -no-link-chkparser/U_binding.o parser/U_constr.o bla.bla.bla... the gmake system gives the error gmake[2]: execvp: ghc-0.29: Arg list too long gmake[2]: *** [hsc] Error 127 Regards, -- Arne Arne John Glenstrup Hitachi Advanced Research Laboratory Hatoyama, Saitama 350-03, Japan tel: +81-492-96-6111 fax: +81-492-96-6006 mailto:[EMAIL PROTECTED] http://www.diku.dk/~panic/
Re: GHC 2.03 info files
First of all a big "Thank you" for all the quick replies! Simon L Peyton Jones wrote: > [...] > http://www.dcs.gla.ac.uk/fp/software/ghc > > For some wierd reason I can't contact Glasgow's web server at the moment. I can't either, but I keep on trying. :-[ BTW, Sigbjorn's ftp-URL doesn't work, either. > [...] Please do let us know what errors you found. I'm embarassed > they happened and would like to fix them. Sorry for my rather short error description, but I'm a bit in a hurry at the moment. Next week I'll mail the whole (sad) story... :-] Sven "Under-pressure" Panne -- Sven PanneTel.: +49/89/2178-2235 LMU, Institut fuer Informatik FAX : +49/89/2178-2211 LFE Programmier- und Modellierungssprachen Oettingenstr. 67 mailto:[EMAIL PROTECTED]D-80538 Muenchen http://www.pms.informatik.uni-muenchen.de/mitarbeiter/panne
incomplete pattern warnings
Minor problem with the check for incomplete patterns: Under ghc-2.02, the following program gives a warning about "Possibly incomplete patterns in the definition of function `f'" even though the patterns clearly are complete. module Test (f) where f (x:_) _ = x f _ (x:_) = x f [] [] = error "both empty" Replacing the second line of f with f [] (x:_) = x eliminates the warning. Since this is a warning rather than an error, and since the warning message does clearly indicate that it might be wrong (by saying "Possibly"), this is merely an annoyance rather than a serious problem. Still, if the compiler is going to get this wrong, you would expect it to be on a harder example than this. Chris
