| Summary: | recent bootstrap.sh build of gcc fails on K6-2 systems | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | D Tuinstra <tuinstra> |
| Component: | [OLD] Core system | Assignee: | Gentoo Linux bug wranglers <bug-wranglers> |
| Status: | RESOLVED INVALID | ||
| Severity: | critical | CC: | cardoe, grandmasterlinux |
| Priority: | High | ||
| Version: | 1.2 | ||
| Hardware: | x86 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
Problem may be due to obscure K6-2 bug: architectural flaw in speculative execution which shows up under extreme CPU stress. See http://membres.lycos.fr/poulot/k6bug.html for details (and thanks to mendezke on the forums). Yesterday's runs had too many lines scrolled off the screen, but two runs today showed Signal 11 halts. Am trying again with open case (in case its heat-based signal 11) and other processes running to possibly lessen chances of speculative execution problem. Will report on results. OK, just got third signal-11, and in yet another place. Based on this, and on discussions in forum during the last day, I believe the problem is due to the fault in some K6-2 chips as described in the link mentioned in previous comment. I'm new to bugzilla and don't know proper etiquette for closing bugs (or even if I have permissions to do so). Can someone please advise, or do it? Many thanks, and sorry for the hassle. I'm trying to install Gentoo on a K6-2 450 as well and it always dies on bzip2's tests. I actually should have a K6-2 450 in my hands in the next couple of days... I can do some tests on this problem soon. I'll post here when the box has arrived and I have started testing. I am very very new to gentoo and have been using linux for a little over a year. I'm still getting the hang of everything, so pardon me for any "weirdness" in my request. Anyway, I just downloaded and installed the stage1 file, got to step 11 and had the same problem and the same file. oh, the download was 7/27/02. Only thing is that I'm running a Pentium 133 chip. I'm not sure if that is Pentium I or Pentium II though. If you need further information let me know, but it sounds like it is the same problem originally mention. Thank you, Jason The real problem with it is the fact that a K6 processor is really more of a RISC processor... it doesn't handle straight up x86 code 100% perfectly. On that box I'm always getting "Internal Compiler Errror!" when I compile stuff. It's just that AMD made a crappy product as far as I know. As far as the Pentium problem. Have you checked your RAM? Well, for those that are curious, I got it to work on the Pentium 133. I did some reading and experimenting and I think the fix was changing my flags to "i586" with "-mcpu" and "-02". I tried cooling the system, slowing it down, and extensive memory and drive testing. As soon as I verify I can reboot (I currently emerging system) I will set my CPU back to normal, change my BIOS back, and try compiling everything again. It does appear with the proper settings everything will work. By the way I have Pentium 133, 64MB RAM, and 2.5G Hardrive. As far as the K6-2's.... check this link out... Explains a bit.. http://membres.lycos.fr/poulot/k6bug_faq.html Since we've obviously determined it's hardware and not Gentoo related.. This one gets the big RESOLVE |
[Note: This problem did not occur about a week ago on a different K6-2 system. This and the error messages (below) puts suspicion on the newly bumped gcc 2.x version.] In following install instructions (www.gentoo.org/doc/build.html) for a K6-2 system, the bootstrap build in step 11 ("Progressing from stage 1 to stage 2) fails. This occurred after doing "emerge rsync" on 5 July 2002, and also "emerge --clean rsync" on 6 July 2002. Prior to running bootstrap.h, /etc/make.conf was edited to uncomment the lines in the "K6 exclusive" section. After cd'ing to /usr/portage and giving the "scripts/bootstrap.sh" command, the bootstrap stopped after about half an hour of configuring and compiling. The error seems to be in gcc-2.95.3/gcc/c-iterate.c, which gives multiple "undefined reference" errors (could it be a missing "include" line in the source file?). The messages still visible on the screen are (transcribing here, omitting redundant err msg detail): ---------------------------------- ... 303: `const_int_rtx' 310: `emit_note' c-iterate.o: In function `iterator_loop_epilogue' 339: `emit_note' 344: `const_int_rtx' 357: `emit_note' collect2: ld returned 1 exit status make[2]: *** [cc1] Error 1 make[2]: Leaving directory `/var/tmp/portage/gcc-2.95.3-r7/work/gcc-2.95/gcc' make[1]: *** [bootstrap-lean] Error 2 make[1]: Leaving directory <same as above> make: *** [bootstrap-lean] Error 2 !!! ERROR: The ebuild did not complete successfully. !!! Function src_compile, Line 33, Exitcode 2 !!! (no error message) !!! emerge aborting on /usr/portage/sys-devel/gcc/gcc-2.95.3-r7.ebuild ---------------------------------- The system is a K6-2/450, but the forum has reports of similar problems on a range of K6-2's (http://forums.gentoo.org/viewtopic.php?t=7153&highlight=k62). The forums also report that commenting out the "K6" lines in /etc/make.conf and trying again doesn't help (I hope this justifies my bumping the severity above "normal") uname -a: Linux (none) 2.4.19-gentoo-r7 #4 Wed Jun 5 07:09:05 Local time zone must be set--see zic manuali586 AuthenticAMD Thanks for your help!