digitalmars.D - gdc / nokia / arm-*-gnueabi
- Michael Dominic Kostrzewa (39/39) Jan 13 2007 Hey
- Frits van Bommel (4/50) Jan 13 2007 A link to that has been posted somewhere in these groups a while back.
- Walter Bright (2/4) Jan 13 2007 Please let me know what those patches are, so they can be folded in.
- John Reimer (10/57) Jan 13 2007 This is great! D will likely prove to be a massive productivity boost f...
- =?ISO-8859-1?Q?Lu=EDs_Marques?= (5/5) Jan 13 2007 Hello Michael,
- Walter Bright (2/4) Jan 13 2007 I agree.
- Kris (5/9) Jan 13 2007 Yeah, me too. The number of computers on desks is dwarfed by the number ...
- Chad J (7/55) Jan 14 2007 I'm sorry I can't really help since compiler crashes scare me, but I
Hey I'm building an arm gdc cross-compiler that'll compile D code to run on maemo[0] platforms (nokia 770[1] and nokia n800[2] internet tablets). There is a bunch of us here at nokia interested in experimenting with D/making it a core part of our platform in the long run. The good: I was able to build an arm-*-linux cross-compiling gdc toolchain & phobos. It works perfectly. The produced code runs correctly on the device, even including the garbagge-collector. I had to apply some patching here and there (mostly to phobos) but nothing specifically magical. Screenshot: http://www.flickr.com/photo_zoom.gne?id=329290303&size=l The bad: We're no longer using arm-*-linux target. We're using the newer arm-*-linux-gnueabi (armel) abi. This is a more powerfull eabi that allows thumb interworking, etc. The gdc doesn't seem to work when compiled for this target. To be precise -- the cc1d segfaults during compilation of certain language structs (it happens when trying to compile phobos). The segfaults are not totally random nor _very_ often (ie. it's possible to compile some phobos files without problems). But something is obviously broken ;) Some data: Debug backtrace of cc1d leading to segfault: http://pastebin.ca/315685 Outout during phobos building: http://pastebin.ca/315687 Strace of cc1d leading to segfault: http://pastebin.ca/315694 I'm building gdc using gcc-3.4.4 from CodeSourcery (this is our default toolchain). Any help/pointers would be appreciated! Regards, [0]: http://www.maemo.org [1]: http://www.nokia.com/770 [2]: http://www.nseries.com/n800 -- Michael Dominic Kostrzewa Nokia - Multimedia Helsinki
Jan 13 2007
Frits van Bommel <fvbommel REMwOVExCAPSs.nl> Michael Dominic Kostrzewa wrote:Hey I'm building an arm gdc cross-compiler that'll compile D code to run on maemo[0] platforms (nokia 770[1] and nokia n800[2] internet tablets). There is a bunch of us here at nokia interested in experimenting with D/making it a core part of our platform in the long run. The good: I was able to build an arm-*-linux cross-compiling gdc toolchain & phobos. It works perfectly. The produced code runs correctly on the device, even including the garbagge-collector. I had to apply some patching here and there (mostly to phobos) but nothing specifically magical. Screenshot: http://www.flickr.com/photo_zoom.gne?id=329290303&size=lA link to that has been posted somewhere in these groups a while back. Good work.The bad: We're no longer using arm-*-linux target. We're using the newer arm-*-linux-gnueabi (armel) abi. This is a more powerfull eabi that allows thumb interworking, etc. The gdc doesn't seem to work when compiled for this target. To be precise -- the cc1d segfaults during compilation of certain language structs (it happens when trying to compile phobos). The segfaults are not totally random nor _very_ often (ie. it's possible to compile some phobos files without problems). But something is obviously broken ;) Some data: Debug backtrace of cc1d leading to segfault: http://pastebin.ca/315685 Outout during phobos building: http://pastebin.ca/315687 Strace of cc1d leading to segfault: http://pastebin.ca/315694 I'm building gdc using gcc-3.4.4 from CodeSourcery (this is our default toolchain). Any help/pointers would be appreciated! Regards, [0]: http://www.maemo.org [1]: http://www.nokia.com/770 [2]: http://www.nseries.com/n800I think D.gnu is a more appropriate group for this. Folluwup set.
Jan 13 2007
Michael Dominic Kostrzewa wrote:I had to apply some patching here and there (mostly to phobos) but nothing specifically magical.Please let me know what those patches are, so they can be folded in.
Jan 13 2007
On Sat, 13 Jan 2007 22:28:16 +0200, Michael Dominic Kostrzewa wrote:Hey I'm building an arm gdc cross-compiler that'll compile D code to run on maemo[0] platforms (nokia 770[1] and nokia n800[2] internet tablets). There is a bunch of us here at nokia interested in experimenting with D/making it a core part of our platform in the long run. The good: I was able to build an arm-*-linux cross-compiling gdc toolchain & phobos. It works perfectly. The produced code runs correctly on the device, even including the garbagge-collector. I had to apply some patching here and there (mostly to phobos) but nothing specifically magical. Screenshot: http://www.flickr.com/photo_zoom.gne?id=329290303&size=l The bad: We're no longer using arm-*-linux target. We're using the newer arm-*-linux-gnueabi (armel) abi. This is a more powerfull eabi that allows thumb interworking, etc. The gdc doesn't seem to work when compiled for this target. To be precise -- the cc1d segfaults during compilation of certain language structs (it happens when trying to compile phobos). The segfaults are not totally random nor _very_ often (ie. it's possible to compile some phobos files without problems). But something is obviously broken ;) Some data: Debug backtrace of cc1d leading to segfault: http://pastebin.ca/315685 Outout during phobos building: http://pastebin.ca/315687 Strace of cc1d leading to segfault: http://pastebin.ca/315694 I'm building gdc using gcc-3.4.4 from CodeSourcery (this is our default toolchain). Any help/pointers would be appreciated! Regards, [0]: http://www.maemo.org [1]: http://www.nokia.com/770 [2]: http://www.nseries.com/n800This is great! D will likely prove to be a massive productivity boost for Nokia developers; I'm sure this will eventually be the case even with third party developers working on Nokia-based projects. It would be great if we could get tango working on this device at some point, as well. Feel free to drop in on the irc chat groups #d and #d.tango (freenode) where you'll likely get all the realtime help you need. All the best! -JJR
Jan 13 2007
Hello Michael, I just wanted to say, please continue this great effort. It will be very appreciated. Best regards, Luís
Jan 13 2007
Luís Marques wrote:I just wanted to say, please continue this great effort. It will be very appreciated.I agree.
Jan 13 2007
== Quote from Walter Bright (newshound digitalmars.com)'s articleLuís Marques wrote:Yeah, me too. The number of computers on desks is dwarfed by the number of "smart" devices. Getting D running on these platforms has always been a primary key to mass adoption, IMO. It's that foot in the door, with a size-287 boot on the end ... Nice work!I just wanted to say, please continue this great effort. It will be very appreciated.I agree.
Jan 13 2007
Michael Dominic Kostrzewa wrote:Hey I'm building an arm gdc cross-compiler that'll compile D code to run on maemo[0] platforms (nokia 770[1] and nokia n800[2] internet tablets). There is a bunch of us here at nokia interested in experimenting with D/making it a core part of our platform in the long run. The good: I was able to build an arm-*-linux cross-compiling gdc toolchain & phobos. It works perfectly. The produced code runs correctly on the device, even including the garbagge-collector. I had to apply some patching here and there (mostly to phobos) but nothing specifically magical. Screenshot: http://www.flickr.com/photo_zoom.gne?id=329290303&size=l The bad: We're no longer using arm-*-linux target. We're using the newer arm-*-linux-gnueabi (armel) abi. This is a more powerfull eabi that allows thumb interworking, etc. The gdc doesn't seem to work when compiled for this target. To be precise -- the cc1d segfaults during compilation of certain language structs (it happens when trying to compile phobos). The segfaults are not totally random nor _very_ often (ie. it's possible to compile some phobos files without problems). But something is obviously broken ;) Some data: Debug backtrace of cc1d leading to segfault: http://pastebin.ca/315685 Outout during phobos building: http://pastebin.ca/315687 Strace of cc1d leading to segfault: http://pastebin.ca/315694 I'm building gdc using gcc-3.4.4 from CodeSourcery (this is our default toolchain). Any help/pointers would be appreciated! Regards, [0]: http://www.maemo.org [1]: http://www.nokia.com/770 [2]: http://www.nseries.com/n800I'm sorry I can't really help since compiler crashes scare me, but I would like to say that I love what you are doing. I hope to write games for mobile devices as soon as college goes away, and D is my language of choice. So the more mobile devices with D support, the better. Thank you so much!
Jan 14 2007