From: Joachim B. <no...@de...> - 2007-09-03 13:23:14
|
Hi, I=E2=80=99m currently getting the gtk2hs package in Debian in shape. Can yo= u comment on this bug: =E2=80=9COn Alpha, I can only handle 32 bytes of non-floating-point argumen= ts to foreign export dynamic=E2=80=9D http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=3D410594 Will gtk2hs be able to run on alpha, or is there no point in trying. If there is, how can we make it compile? Thanks, Joachim --=20 Joachim "nomeata" Breitner Debian Developer no...@de... | ICQ# 74513189 | GPG-Keyid: 4743206C JID: joa...@am... | http://people.debian.org/~nomeata |
From: Duncan C. <dun...@wo...> - 2007-09-04 12:41:05
|
On Mon, 2007-09-03 at 14:58 +0200, Joachim Breitner wrote: > Hi, >=20 > I=E2=80=99m currently getting the gtk2hs package in Debian in shape. Can = you > comment on this bug: > =E2=80=9COn Alpha, I can only handle 32 bytes of non-floating-point argum= ents to > foreign export dynamic=E2=80=9D > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=3D410594 >=20 > Will gtk2hs be able to run on alpha, or is there no point in trying. If > there is, how can we make it compile? Probably only by just not compiling those functions, using cpp. :-( So finding out which ones don't work on alpha would involve having shell access to an alpha box and just iterating with make and adding {--} around the unusable ones (and keeping a list). Not much good if you don't have direct shell access to the build machine (which I think you don't right?) Duncan |
From: Axel S. <A....@ke...> - 2007-09-04 12:41:47
|
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Joachim, On Sep 3, 2007, at 13:58, Joachim Breitner wrote: > Hi, > > I=1B$B!G=1B(Bm currently getting the gtk2hs package in Debian in = shape. Can you > comment on this bug: > =1B$B!H=1B(BOn Alpha, I can only handle 32 bytes of non-floating-point = =20 > arguments to > foreign export dynamic=1B$B!I=1B(B > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=3D410594 > > Will gtk2hs be able to run on alpha, or is there no point in =20 > trying. If > there is, how can we make it compile? As it stands, there's no point trying. This is a GHC limitation. A =20 similar limitation was present on Sparc, but they've fixed it. The =20 reason is AFAIK that if you pass more than 32 bytes worth of non-=20 floating-point arguments, the arguments cannot be passed in registers =20= but are spilled onto the stack. GHC simply does not implement the =20 code to deal with this situation. There are two ways to fix this: a) add the missing code in GHC b) conditionally omit the functions in gtk2hs that use more that 32-=20 bytes in callback functions We did b) for Sparc up to some point. But since it cripples Gtk2Hs =20 somewhat it is not a nice solution. I think somebody should file a bug for ghc and then link the Gtk2Hs =20 bug to the ghc bug. Thanks for trying! Axel. > Thanks, > Joachim > --=20 > Joachim "nomeata" Breitner > Debian Developer > no...@de... | ICQ# 74513189 | GPG-Keyid: 4743206C > JID: joa...@am... | http://people.debian.org/~nomeata > > ----------------------------------------------------------------------=20= > --- > This SF.net email is sponsored by: Splunk Inc. > Still grepping through log files to find problems? Stop. > Now Search log events and configuration files using AJAX and a =20 > browser. > Download your FREE copy of Splunk now >> http://get.splunk.com/=20 > _______________________________________________ > Gtk2hs-devel mailing list > Gtk...@li... > https://lists.sourceforge.net/lists/listinfo/gtk2hs-devel -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (Darwin) iD8DBQFG3VIoA+OhltVdZ9oRArafAKCAKW6zSDM0uAcO2szDbXeRkMCyiwCdHP0r 3j9lHl4dpCZUTRMwu1mQHok=3D =3D9FYL -----END PGP SIGNATURE----- |
From: Joachim B. <no...@de...> - 2007-09-04 13:03:10
|
Hi, Am Dienstag, den 04.09.2007, 13:40 +0100 schrieb Axel Simon: > On Sep 3, 2007, at 13:58, Joachim Breitner wrote: > > I=E2=80=99m currently getting the gtk2hs package in Debian in shape. Ca= n you > > comment on this bug: > > =E2=80=9COn Alpha, I can only handle 32 bytes of non-floating-point =20 > > arguments to > > foreign export dynamic=E2=80=9D > > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=3D410594 > > > > Will gtk2hs be able to run on alpha, or is there no point in =20 > > trying. If > > there is, how can we make it compile? >=20 > As it stands, there's no point trying. This is a GHC limitation. A =20 > similar limitation was present on Sparc, but they've fixed it. The =20 > reason is AFAIK that if you pass more than 32 bytes worth of non-=20 > floating-point arguments, the arguments cannot be passed in registers =20 > but are spilled onto the stack. GHC simply does not implement the =20 > code to deal with this situation. There are two ways to fix this: >=20 > a) add the missing code in GHC > b) conditionally omit the functions in gtk2hs that use more that 32-=20 > bytes in callback functions >=20 > We did b) for Sparc up to some point. But since it cripples Gtk2Hs =20 > somewhat it is not a nice solution. >=20 > I think somebody should file a bug for ghc and then link the Gtk2Hs =20 > bug to the ghc bug. >=20 > Thanks for trying! > Axel. So for the gtk2hs package I guess I should just remove alpha from the list of supported architectures with the next upload, right? Greetings, Joachim --=20 Joachim "nomeata" Breitner Debian Developer no...@de... | ICQ# 74513189 | GPG-Keyid: 4743206C JID: joa...@am... | http://people.debian.org/~nomeata |
From: Axel S. <A....@ke...> - 2007-09-04 13:25:32
|
On Sep 4, 2007, at 14:03, Joachim Breitner wrote: > > So for the gtk2hs package I guess I should just remove alpha from the > list of supported architectures with the next upload, right? Yes, that's the only choice for now. Axel. |