From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows 98) Description of problem: Problems compiling on RH 7.1 (Maybe glibc problem ?? ) I experience following problem : gcc2.96-81 crash while compiling kernel. (Segmentation fault). If I reboot and "make" (with or without prior issue of "make clean" ) again,then it crashes at the same line..... Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: Standard kernel compiling procedure. make menuconfig make dep make bzImage... Actual Results: Internal error: Segmentation fault. Please submit a full bug report. ****** please read Additional information about this bug !!! Expected Results: compiling working kernel Additional Information: I upgraded gcc to 2.96-85(As recomended solution for this bug 31168 in "bugzilla" ) , but problem persists. I make some examination of this, beause just few hours later I compiled the same kernel souce code on the same machine (Machine K6-2_3D/300, 96MB Ram, 255MB swap space,in most cases, there is no more swap space used than 128 kb. ) but runing RH 7.0,kernel 2.4.10 with absolutely no problems. I increased Ram to 128 (brand new dimm module) and added another 64 MB to swap space to avoid memory hw fail + kernel memory/swap problems. NO effect. For the same reason I tried to boot machine with different kernels I have (2.4.10 , 2.4.3(from RH) ,2.4.16 and 2.2.16-22(From RH) ) and compile again.No effect. I downgraded GCC to 2.96-54 ,the same version as on the previous instalation,( where it runs without problems )... No effect. If I change the glibc from 2.2.2-10 to 2.2.4-19.3 for RH7.2 ( rpm --nodeps...), it works in many cases, gcc makes segfault only occasionaly and after segfault I can run "make" again and continue with compilation (Part of source code,that cause segfault now compiles OK ) and It produce working kernel(!)..... If I downgrade glibc to 2.2-5 ("rpm --upgrade --oldpackage --force -- replacefiles"), everything works fine.(I try to make full compilation procedure "make clean menuconfig dep bzImage modules" 4 times without an error).It work ifne even with gcc upgradet to gcc 2.96-85... So, there must be something nasty in gcc 2.2.2-10 and above........ Solution: downgrade glibc to 2.2.5-10 PS: downgrading glibc also solved problem with long execution time/failing/crashing/freezing of "ssh-keygen" program.....
sorry, misttype ."So, there must be something nasty in gcc 2.2.2-10 and above........" should speak about glibc 2.2.2-10 1
sorry, misttype ."So, there must be something nasty in gcc 2.2.2-10 and above........" should speak about glibc 2.2.2-10 and "downgrade glibc to 2.2.5- 10" should speak about "glibc 2.2-5"
Since glibc-2.2.4-19.3 is the official errata glibc for 7.1, I think there is no point to talk about glibc-2.2.2-10. And if you get random non-reproduceable segfaults, it doesn't smell like gcc or glibc bug.
Random non reproducable segfaults smell like hw/memory problems. But: With glibc-2.2.2-10. - Segfault ALWAYS (At the same sourcecode line) and unable to continue work (ie building a working kernel) With RedHat official errarta glibc-2.2.4-19.3 Segfault randomly. With old glibc-2.2-5 NO SEGFAULTS and everything works OK.I made intensive tests with this library to be sure, that it work well.(include 4 cycles of standard kernel compiling procedure "make clean make deep bzImage modules" It takes almost all day) I do not thing, that ancient glibc could solve any hardware problem. From my point of wiew, it smell like:glibc-2.2.2-10. introduced some mysterious bug, that was not present in the glibc-2.2-5 and in the relase glibc-2.2.4-19.3 was this bug partially fixed. Any idea what/how to make future tests to track this down ? PS: Execuse my wrong english.
Additional info: gcc 2.96-85 + glibc-2.2.4-19.3 work OK (no segfault) if gcc used without -pipe option (ie comment out "CFLAGS += -pipe "in /arch/i386 while compilling kernel)
Closing as a two years old and non-reproducable.