Menu
▾
▴
colinux-users
|
From: Ronald P. <pij...@ds...> - 2004-08-19 19:34:42
|
Hi all, This week I installed snapshot 20040710. I ran into a couple of problems: - When logging in (vnc + XDMCP) WindowMaker now gets a floating point exception. This doesn't happen with the previous snapshots or in native linux. - /dev/cobd1 is mapped to the first partition, which holds a ext3 fs. Until now I could mount it with no problems. However, in this snapshot I cannot anymore: "no such device". Just my two cents... Oh, by the way, I saw some announcements of "other" snapshots in the mailing list. What is the status of these snapshots compared with the ones on the official site (colinux.org/snapshots) ? The latest snapshot seems quite old. Regards, Ronald. |
|
From: Peng <rin...@gm...> - 2004-08-19 20:10:49
|
Which distribution are you using? On Thu, 19 Aug 2004 21:34:35 +0200, Ronald Pijnacker <pij...@ds...> wrote: > Hi all, > > This week I installed snapshot 20040710. I ran into a couple of problems: > > - When logging in (vnc + XDMCP) WindowMaker now gets a floating point > exception. This doesn't happen with the previous snapshots or in > native linux. > > - /dev/cobd1 is mapped to the first partition, which holds a ext3 fs. > Until now I could mount it with no problems. However, in this snapshot > I cannot anymore: "no such device". >W > Just my two cents... > > Oh, by the way, I saw some announcements of "other" snapshots in the mailing > list. What is the status of these snapshots compared with the ones on the > official site (colinux.org/snapshots) ? The latest snapshot seems quite old. > > Regards, > > Ronald. > > ------------------------------------------------------- > SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media > 100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33 > Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift. > http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285 > _______________________________________________ > coLinux-users mailing list > coL...@li... > https://lists.sourceforge.net/lists/listinfo/colinux-users > |
|
From: Ronald P. <pij...@ds...> - 2004-08-21 11:49:27
|
On Thu, Aug 19, 2004 at 04:10:37PM -0400, Peng wrote: > Which distribution are you using? I am using the same Gentoo linux that I also use natively. I do compile with optimization flags for the Pentium4 that I have, but that's never given any problems before. Regards, Ronald. |
|
From: Peng <rin...@gm...> - 2004-08-21 19:53:44
|
The floating point error is probably caused by the CFLAGS. I solved it by recompiling the system (at least, xfree) with "-march=i686". Good luck, Peng On Sat, 21 Aug 2004 13:49:20 +0200, Ronald Pijnacker <pij...@ds...> wrote: > On Thu, Aug 19, 2004 at 04:10:37PM -0400, Peng wrote: > > Which distribution are you using? > > I am using the same Gentoo linux that I also use natively. > I do compile with optimization flags for the Pentium4 that I have, but > that's never given any problems before. > > Regards, > > Ronald. > |
|
From: <sl...@bl...> - 2004-08-21 20:34:36
|
Peng <rin...@gm...> writes: > The floating point error is probably caused by the CFLAGS. I solved > it by recompiling the system (at least, xfree) with "-march=i686". By the way, this is equivalent to -march=pentiumpro. Inside GCC, using -march=pentiumpro instead of -march=pentium4 uses the PROCESSOR_PENTIUMPRO flag instead of PROCESSOR_PENTIUM4, and turns off PTA_SSE, PTA_SSE2, PTA_MMX, PTA_PREFETCH_SSE. (Both -march=pentium2 and -march=pentium3 also use PROCESSOR_PENTIUMPRO.) I am making a wild guess that it will be enough to turn off PTA_MMX. I want to try some recompiling with "-march=pentium4 -mno-mmx" to see if that is enough to fix things. -- Joe > Good luck, > > Peng > > On Sat, 21 Aug 2004 13:49:20 +0200, Ronald Pijnacker <pij...@ds...> wrote: > > On Thu, Aug 19, 2004 at 04:10:37PM -0400, Peng wrote: > > > Which distribution are you using? > > > > I am using the same Gentoo linux that I also use natively. > > I do compile with optimization flags for the Pentium4 that I have, but > > that's never given any problems before. > > > > Regards, > > > > Ronald. > > > > > ------------------------------------------------------- > SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media > 100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33 > Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift. > http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285 > _______________________________________________ > coLinux-users mailing list > coL...@li... > https://lists.sourceforge.net/lists/listinfo/colinux-users |
|
From: <sl...@bl...> - 2004-08-19 20:41:51
|
Ronald Pijnacker <pij...@ds...> writes: > This week I installed snapshot 20040710. I ran into a couple of problems: > > - When logging in (vnc + XDMCP) WindowMaker now gets a floating point > exception. This doesn't happen with the previous snapshots or in > native linux. Can you check with gdb if it is failing for the same reason as Firefox and Xdvi fail for me? See http://www.macs.hw.ac.uk/~jbw/colinux-floating-point.mbox for my messages containing details of how to check a core dump with gdb. If you are seeing the same kind of failure I am seeing, then some instruction to transfer a value into a floating point register will have yielded a wacky value in the register, perhaps a few instructions before the crash. For me, I always see the programs crash when they use cvttsd2si to convert the content of a floating point register to an integer, but it seems something has already gone wrong a few steps before the crash. Also, what precise CPU do you have? I have an Intel Pentium M (mobile) CPU. > - /dev/cobd1 is mapped to the first partition, which holds a ext3 fs. > Until now I could mount it with no problems. However, in this snapshot > I cannot anymore: "no such device". If you are using devfs (e.g., if you are using Gentoo), then the pseudo file /dev/cobd1 will not be created. Instead, devfs will create /dev/cobd/1. See the earlier messages on the mailing list about this. -- Joe |
|
From: Henry N. <Hen...@ar...> - 2004-08-21 12:42:13
|
Ronald Pijnacker wrote: > Hi all, > > This week I installed snapshot 20040710. I ran into a couple of problems: > > - When logging in (vnc + XDMCP) WindowMaker now gets a floating point > exception. This doesn't happen with the previous snapshots or in > native linux. > > - /dev/cobd1 is mapped to the first partition, which holds a ext3 fs. > Until now I could mount it with no problems. However, in this snapshot > I cannot anymore: "no such device". > > Just my two cents... > > Oh, by the way, I saw some announcements of "other" snapshots in the mailing > list. What is the status of these snapshots compared with the ones on the > official site (colinux.org/snapshots) ? The latest snapshot seems quite old. > > Regards, > > Ronald. The official snapshot is completly. Some newer version works, but they have no installer. You must install the last snapshoot, than download the executables, kernel, Modules, extract this into your favorite directory and call a simply colinux-deamon.exe --remove-driver colinux-deamon.exe --install-driver After this you can test some of the non official versions. http://home.arcor.de/henryn/colinux/colinux-0.6.2-20040817.zip http://home.arcor.de/henryn/colinux/modules-0.6.2-20040817.tar.gz Please try this, think works better. If you interrest the difference of this version, you should install monotone and get the log of the source. Henry |
|
From: Henry N. <Hen...@ar...> - 2004-08-21 22:55:16
|
Joe Wells (reverse mailbox letters to reply) wrote: > Peng <rin...@gm...> writes: > > >>The floating point error is probably caused by the CFLAGS. I solved >>it by recompiling the system (at least, xfree) with "-march=i686". > > > By the way, this is equivalent to -march=pentiumpro. Inside GCC, > using -march=pentiumpro instead of -march=pentium4 uses the > PROCESSOR_PENTIUMPRO flag instead of PROCESSOR_PENTIUM4, and turns off > PTA_SSE, PTA_SSE2, PTA_MMX, PTA_PREFETCH_SSE. (Both -march=pentium2 > and -march=pentium3 also use PROCESSOR_PENTIUMPRO.) > > I am making a wild guess that it will be enough to turn off PTA_MMX. > I want to try some recompiling with "-march=pentium4 -mno-mmx" to see > if that is enough to fix things. > I'm view into compile scripts. Kernel was compiled for i586. Show into conf/linux-config: CONFIG_M586=y # CONFIG_M586MMX is not set # CONFIG_M686 is not set # CONFIG_MPENTIUMII is not set # CONFIG_MPENTIUMIII is not set # CONFIG_MPENTIUMM is not set # CONFIG_MPENTIUM4 is not set ... But the colinux libraries and daemons will compile for 686. See into build-common.sh: TARGET=i686-pc-mingw32 This compile the cross compiler for i686 and all targets was compile for i686, also if the build do no set any CFLAGS. Is this your problem? If yes, should we compile all for i586 only? I'm using Pentuim4 and do not get any errors. Henry |
|
From: <sl...@bl...> - 2004-08-30 12:00:16
|
Henry Nestler <Hen...@ar...> writes: > Joe Wells wrote: > > > Peng <rin...@gm...> writes: > > > >>The floating point error is probably caused by the CFLAGS. I solved > >>it by recompiling the system (at least, xfree) with "-march=i686". > > > > By the way, this is equivalent to -march=pentiumpro. Inside GCC, > > using -march=pentiumpro instead of -march=pentium4 uses the > > PROCESSOR_PENTIUMPRO flag instead of PROCESSOR_PENTIUM4, and turns off > > PTA_SSE, PTA_SSE2, PTA_MMX, PTA_PREFETCH_SSE. (Both -march=pentium2 > > and -march=pentium3 also use PROCESSOR_PENTIUMPRO.) > > > > I am making a wild guess that it will be enough to turn off PTA_MMX. > > I want to try some recompiling with "-march=pentium4 -mno-mmx" to see > > if that is enough to fix things. > > I'm view into compile scripts. Kernel was compiled for i586. > Show into conf/linux-config: > CONFIG_M586=y > # CONFIG_M586MMX is not set > # CONFIG_M686 is not set > # CONFIG_MPENTIUMII is not set > # CONFIG_MPENTIUMIII is not set > # CONFIG_MPENTIUMM is not set > # CONFIG_MPENTIUM4 is not set > ... > > But the colinux libraries and daemons will compile for 686. > See into build-common.sh: > TARGET=i686-pc-mingw32 > This compile the cross compiler for i686 and all targets was compile > for i686, also if the build do no set any CFLAGS. > > Is this your problem? If yes, should we compile all for i586 only? I don't know if it is currently making any difference that the *kernel* is compiled for 586. I think compiling the kernel for 586 is probably conservative enough. It could conceivably make a difference that the coLinux libraries are compiled for 686. What I was proposing above is to recompile some *applications* with use of the MMX features turned off. What is needed is for there to be a person who is having the floating point problems with their machine who is also good at rebuilding/debugging coLinux and the Linux kernel. :-( > I'm using Pentuim4 and do not get any errors. Are you using the Pentium M? (The "M" is for mobile.) -- Joe |
|
From: <sl...@bl...> - 2004-09-13 09:29:17
|
Henry Nestler <Hen...@ar...> writes: > Joe Wells wrote: > > > Are you using the Pentium M? (The "M" is for mobile.) > > Pentium 4M. Yes, it's mobile :-) I don't know if this matters so much now that the floating point exception problem is supposedly solved (will test later today), but a Pentium 4M is a very different CPU from a Pentium M. The Pentium M series implements the same instruction set as the Pentium 4 series, but is internally designed in radically different way. I don't know whether this has anything to do with why you couldn't reproduce the floating point bug. > You can self recompiling all executable and kernel. Maybe someday. If this latest coLinux snapshot fixes my problems, then I hope I will not have to. -- Joe |
Thanks for helping keep SourceForge clean.
X