Return to BSD News archive
Newsgroups: comp.unix.bsd Path: sserve!manuel.anu.edu.au!munnari.oz.au!spool.mu.edu!sgiblab!tsoft!barfeau From: bbs.barfeau@tsoft.sf-bay.org (David Fox) Subject: Re: running out of memory in compiling kernel Message-ID: <qT42sB1w165w@tsoft.sf-bay.org> Sender: bbs@tsoft.sf-bay.org (BBS User) Organization: The TSoft BBS and Public Access Unix, +1 415 969 8238 Date: Thu, 22 Oct 1992 17:30:37 GMT Lines: 38 David.Fox@Saigon.COM wrote: : Hi bsders, : : I got the kernel source code and started to compile the kernel from scratch. : Everything was seemingly going smoothly until I noticed that the make was : stuck trying to compile nfs/nfs_serv.c, and it was taking about 1 1/2 HOURS : to compile this particular file. After this long time, I decided to try : >I've been meaning to post about this too. What I find is that when i rebuild >from scratch (i.e. after deleting all objs) the compiler eventually hits >what seems to be an infinite loop. If I control-C out of make, and >then make again (restarting a compile of the same file) it once again >compiles forever. Re-booting however and doing make, the offending >file will re-compile in a few seconds. However, usually later on >the same thing happens again with a different file. Again a re-boot >fixes the problem. My guess as to what is happening is that >there is a memory leak somewhere which causes the compiler to hang >after doing a certain amount of work. Re-booting restores the memory >(or swap space?) and allows more compiling to happen until the memory >(or swap space?) fills up again. I have only seen this when re-compiling >the kernel, but I haven't built anything else with that many files to >compile all at once. After experimenting some more, it definitely IS a memory leak in the make program. I was recompiling the kernel last night, and periodically did a suspend followed by a ps -alx. The VSZ kept increasing until the memory consumed by make was over one megabyte, and the memory used by cc1 was at almost two megabytes before the system became unusable. -- David Fox (bbs.barfeau@tsoft.sf-bay.org)