Date: Wed, 29 Dec 2004 00:04:18 EST-10EDT,10,-1,0,7200,3,-1,0,7200,3600 Subject: [UnixOS2_Archive] No. 487 ************************************************** Tuesday 28 December 2004 Number 487 ************************************************** Subjects for today 1 Re: Fork and AMD K6-2 CPU : lsunley at mb.sympatico.ca 2 Re: Fork and AMD K6-2 CPU : John Poltorak 3 Re: POVRAY : Christian Hennecke" 4 Re: POVRAY : John Poltorak 5 Re: Fork and AMD K6-2 CPU : billn 6 Re: Fork and AMD K6-2 CPU : Dave Yeo" 7 Re: Fork and AMD K6-2 CPU : lsunley at mb.sympatico.ca 8 Re: Fork and AMD K6-2 CPU : Dave Yeo 9 teamos2.ru : John Poltorak **= Email 1 ==========================** Date: Mon, 27 Dec 2004 09:17:31 -0500 From: lsunley at mb.sympatico.ca Subject: Re: Fork and AMD K6-2 CPU In <20041227122639.A80 at warpix.org>, on 12/27/04 at 12:26 PM, John Poltorak said: >On Sun, Dec 26, 2004 at 10:00:37AM -0500, lsunley at mb.sympatico.ca wrote: >> Hi, >> >> Here is the stupid fork() testcase >> >> compile it with gcc -Zomf testfork.c >> >> and run it... >> >> I included the built version that runs on my system. >I just ran it on my system with the AMD-K2 (with the latest kernel) :- >Starting program pid=58 >Process terminated by SIGILL - pid=0x3a - tid=0x01 - >C:\USR\CMDLIB\TESTFORK.EXE Yes, that is the FXSAVE instruction trap. Knut confirmed that the instruction was the one causing the traps on the Innotek support forums. I had posted Dave Yeo's debug analysis there. >On my Thinkpad 600 I get:- >Starting program pid=10359 >Fork started pid=10360 >main line in pid=10359 >Fork 2 started pid=10361 >Fork 2 main line in pid=10361 >Fork 2 main line in pid=10360 That is what is supposed to happen. You can use postgreSQL on your Thinkpad but not on the other machine. > >> Thanks for the help >> >> Lorne >> >> In <20041226071856.22DD5BAA35 at generation.lgisp.net>, on 12/25/04 >> at 11:18 PM, "Dave Yeo" said: >> >> >On Sat, 25 Dec 2004 18:03:35 -0500, lsunley at mb.sympatico.ca wrote: >> >> >>Does anyone out there running programs using fork() have >> >>an AMD K6-2 CPU. I have a case with a SIGILL (illegal instruction) >> >>happening with postgreSQL when it forks. >> >> >> >>The initdb works (which calls the database engine in standalone mode) but >> >>using postmaster (which uses fork() ) fails with a SIGILL >> >> >I have a K2 (300 Mhz) and have had problems with fork() failing on >> >imake.exe (SIGILL). I just changed it to spawn. Unluckily I don't really >> >have the bandwidth right now to DL postgreSQL to test it but if you could >> >come up with a test case I could try it. Dave -- ----------------------------------------------------------- lsunley at mb.sympatico.ca ----------------------------------------------------------- **= Email 2 ==========================** Date: Mon, 27 Dec 2004 16:50:55 +0000 From: John Poltorak Subject: Re: Fork and AMD K6-2 CPU On Mon, Dec 27, 2004 at 09:17:31AM -0500, lsunley at mb.sympatico.ca wrote: > In <20041227122639.A80 at warpix.org>, on 12/27/04 > at 12:26 PM, John Poltorak said: > > >I just ran it on my system with the AMD-K2 (with the latest kernel) :- > > >Starting program pid=58 > > >Process terminated by SIGILL - pid=0x3a - tid=0x01 - > >C:\USR\CMDLIB\TESTFORK.EXE > > Yes, that is the FXSAVE instruction trap. Knut confirmed that the > instruction was the one causing the traps on the Innotek support forums. I > had posted Dave Yeo's debug analysis there. > > >On my Thinkpad 600 I get:- > > >Starting program pid=10359 > >Fork started pid=10360 > >main line in pid=10359 > >Fork 2 started pid=10361 > >Fork 2 main line in pid=10361 > >Fork 2 main line in pid=10360 > > That is what is supposed to happen. You can use postgreSQL on your > Thinkpad but not on the other machine. That's a pity since I'm hoping to set up my PC730 as dedicated test system, but I keep getting all sorts of errors I'm not used to even with emx/gcc 2.8.1. I get lots of errors even when building Perl with a tried and trusted environment and am at something of a loss as to understand why all these SYS 3175 and 3176 errors suddenly occur. Could they be due to the processor in use? Incidentally there are some jumper switches on the processor. Could their settings be responsible for such errors? > -- > ----------------------------------------------------------- > lsunley at mb.sympatico.ca > ----------------------------------------------------------- > -- John **= Email 3 ==========================** Date: Mon, 27 Dec 2004 18:19:49 +0100 (CET) From: "Christian Hennecke" Subject: Re: POVRAY On Mon, 27 Dec 2004 12:10:36 +0000, John Poltorak wrote: >> > Has anyone tried building a recent version of POVRAY on OS/2? >> >> I had a go at it probably six months or a year ago but I don't remember the >> version number. As best I can remember, getting it to compile wasn't very >> difficult but it crashed when I tried to run it and I never got around to >> finding out what was wrong. > >Did the last OS/2 version come with any patches? No. >Wonder what happened to the OS/2 porter... Alexander Mai hasn't posted on the XFree86/OS2 list and also hasn't released any ports for a long time. IIRC he said that he was busy with other things now. Christian Hennecke **= Email 4 ==========================** Date: Mon, 27 Dec 2004 17:36:15 +0000 From: John Poltorak Subject: Re: POVRAY On Mon, Dec 27, 2004 at 06:19:49PM +0100, Christian Hennecke wrote: > On Mon, 27 Dec 2004 12:10:36 +0000, John Poltorak wrote: > > >> > Has anyone tried building a recent version of POVRAY on OS/2? > >> > >> I had a go at it probably six months or a year ago but I don't remember the > >> version number. As best I can remember, getting it to compile wasn't very > >> difficult but it crashed when I tried to run it and I never got around to > >> finding out what was wrong. > > > >Did the last OS/2 version come with any patches? > > No. > > >Wonder what happened to the OS/2 porter... > > Alexander Mai hasn't posted on the XFree86/OS2 list and also hasn't > released any ports for a long time. IIRC he said that he was busy with > other things now. If he's moved on then someone ought to grab his OS/2 pages here:- http://www.lesstif.org/~amai/os2/html/ before they disappear... On his POVRAY page:- http://www.lesstif.org/~amai/os2/html/povray.html he says:- There are other builds of Povray for OS/2 available, but my are usually the most recent ones. In fact it is trivial to create them, since the source is very portable and compiles out of the box as of 3.1x. > Christian Hennecke -- John **= Email 5 ==========================** Date: Mon, 27 Dec 2004 09:52:46 -0800 From: billn Subject: Re: Fork and AMD K6-2 CPU John wrote: > "That's a pity since I'm hoping to set up my PC730 as dedicated test > system, but I keep getting all sorts of errors I'm not used to even with > emx/gcc 2.8.1. I get lots of errors even when building Perl with a tried > and trusted environment and am at something of a loss as to understand why > all these SYS 3175 and 3176 errors suddenly occur. Could they be due to > the processor in use? Incidentally there are some jumper switches on the > processor. Could their settings be responsible for such errors?" Jumper switches are usually for memory speed and processor emulation. Try a google search for the specific upgrade chip for specs. BTW, there is very little reason to hold on to older systems at this point. For those with an ATX box, this is a killer $50 MB & CPU upgrade for old systems: * Processor: AMD duron 1.6GHz+ Bare CPU * Chipset: VIA KT266A * Processor Support: Socket A 266/200MHz Front Side Bus CPU * Memory Support: Up to 2GB PC2400 DDR * Expansion Slot: 1x AGP 4X, 5x PCI, 1x CNR * Other Features: Audio, 10/100 LAN, USB 2.0, For details: http://shop1.outpost.com/ and search for Outpost #: 4296754 For John, I have an older Duron 1.3GHz, MB and 512 MB SDRAM if you have an ATX box to put it in. Includes good heatsink & fan. Shuttle AK32 MB can also take DDR and cpus up to 2400XP. If John can't use it, maybe someone else can. BillN **= Email 6 ==========================** Date: Mon, 27 Dec 2004 12:06:20 -0800 From: "Dave Yeo" Subject: Re: Fork and AMD K6-2 CPU On Mon, 27 Dec 2004 16:50:55 +0000, John Poltorak wrote: >That's a pity since I'm hoping to set up my PC730 as dedicated test >system, but I keep getting all sorts of errors I'm not used to even with >emx/gcc 2.8.1. I get lots of errors even when building Perl with a tried >and trusted environment and am at something of a loss as to understand why >all these SYS 3175 and 3176 errors suddenly occur. Could they be due to >the processor in use? Incidentally there are some jumper switches on the >processor. Could their settings be responsible for such errors? > You could always try underclocking your processer (or/and memory) I had one machine with a bad CPU. Caused Netscape to continually give SYS3175s and my suond driver to act weird. Also a couple of times my CPU fan failed also causing all kinds of weird errors. This machine I'm using came with some 32 MB simms. I plugged in a 128 MB simm and the had to change a few memory related BIOS settings to make it stable. Later I removed the 32 MB simms and stuck in a 256 MB. Now I can run the memory at full speed with 4 way interleaving. Check for over heating, then perhaps try underclocking Dave **= Email 7 ==========================** Date: Mon, 27 Dec 2004 11:11:41 -0500 From: lsunley at mb.sympatico.ca Subject: Re: Fork and AMD K6-2 CPU In <20041227165054.B80 at warpix.org>, on 12/27/04 at 04:50 PM, John Poltorak said: >On Mon, Dec 27, 2004 at 09:17:31AM -0500, lsunley at mb.sympatico.ca wrote: >> In <20041227122639.A80 at warpix.org>, on 12/27/04 >> at 12:26 PM, John Poltorak said: >> >> >I just ran it on my system with the AMD-K2 (with the latest kernel) :- >> >> >Starting program pid=58 >> >> >Process terminated by SIGILL - pid=0x3a - tid=0x01 - >> >C:\USR\CMDLIB\TESTFORK.EXE >> >> Yes, that is the FXSAVE instruction trap. Knut confirmed that the >> instruction was the one causing the traps on the Innotek support forums. I >> had posted Dave Yeo's debug analysis there. >> >> >On my Thinkpad 600 I get:- >> >> >Starting program pid=10359 >> >Fork started pid=10360 >> >main line in pid=10359 >> >Fork 2 started pid=10361 >> >Fork 2 main line in pid=10361 >> >Fork 2 main line in pid=10360 >> >> That is what is supposed to happen. You can use postgreSQL on your >> Thinkpad but not on the other machine. >That's a pity since I'm hoping to set up my PC730 as dedicated test >system, but I keep getting all sorts of errors I'm not used to even with >emx/gcc 2.8.1. I get lots of errors even when building Perl with a tried >and trusted environment and am at something of a loss as to understand >why all these SYS 3175 and 3176 errors suddenly occur. Could they be due >to the processor in use? Incidentally there are some jumper switches on >the processor. Could their settings be responsible for such errors? The SYS3176 errors are illegal opcode errors. It could be that the runtime code uses a number of instructions that are not available on the K6-2. If the code you generate is causing the errors ou could try using the -march option on the compiler to generate K6-2 code. -- ----------------------------------------------------------- lsunley at mb.sympatico.ca ----------------------------------------------------------- **= Email 8 ==========================** Date: Mon, 27 Dec 2004 15:42:48 -0800 From: Dave Yeo Subject: Re: Fork and AMD K6-2 CPU On Mon, 2004-27-12 at 11:11 -0500, lsunley at mb.sympatico.ca wrote: > The SYS3176 errors are illegal opcode errors. It could be that the > runtime > code uses a number of instructions that are not available on the K6-2. > If > the code you generate is causing the errors ou could try using the > -march > option on the compiler to generate K6-2 code. On my K6-2 the only sigills I've gotten seem all to be related to fork. It seems the FXSAVE instruction wasn't introduced until about the Pentium II and a half (350 Mhz Deschutes (sp?)) so quite a few older processors are going to get sigills. Seems that FSAVE would be a better choice though it doesn't save quite as much in one instruction and is possibly slower. I wonder if GCC added the FXSAVE or if Knut used Assembly? Dave **= Email 9 ==========================** Date: Tue, 28 Dec 2004 10:33:35 +0000 From: John Poltorak Subject: teamos2.ru teamos2.ru has quite a number of OS/2 ports of Unix programs as you can see if you look here:- http://teamos2.ru/files/ Unfortunately there are never any patch files so it is impossible to rebuild them and bring them upto date. If anyone on this list is a member of teamos2.ru or knows anyone who is do you know if it is possible to get hold of patches and build instructions? -- John