I get an error when compiling dev-libs/boost-1.34.1-r1 (upgrading from 1.33.1-r1). It failed twice at the same spot with -j4. I tried it with -j1 and it seems to be doing just fine -- it passed the spot where it failed with -j4.
Created attachment 142900 [details] dev-libs:boost-1.34.1-r2:20080207-134141.log There you go -- the build log.
It appears that the compiler crashed. I'm clueless as what might cause it as I'm unable to reproduce this. Is your RAM damaged?
(In reply to comment #2) > It appears that the compiler crashed. I'm clueless as what might cause it as > I'm unable to reproduce this. Is your RAM damaged? > Ooops, maybe I ran out of memory? Though I don't see any oom-killer killing gcc messages in the log.
(In reply to comment #3) > (In reply to comment #2) > > It appears that the compiler crashed. I'm clueless as what might cause it as > > I'm unable to reproduce this. Is your RAM damaged? > > > > Ooops, maybe I ran out of memory? Though I don't see any oom-killer killing gcc > messages in the log. > Physically damaged. When data stored has changed next time it is read. There's a tool named memtest86+ which comes with the minimal install CD that can test your memory for data corruption.
How many cores do you have? And does it work with -j2?
(In reply to comment #5) > How many cores do you have? And does it work with -j2? 2 CPUs x 2 cores each = 4 total. It did work with -j1. I just tried it with -j2 and it also worked. I was also monitoring the free memory and saw it go down from 980M to 200M at times... I guess that when you throw two more parallel jobs in the mix it's not that weird to run out of memory (I have no swap). (In reply to comment #4) > Physically damaged. When data stored has changed next time it is read. There's > a tool named memtest86+ which comes with the minimal install CD that can test > your memory for data corruption. I did test my RAM a couple of months ago and it's ECC RAM that I have so that shouldn't be the case. I just found out that I should compile the kernel with CONFIG_EDAC in order to actually see if everything's alright.
Sorry for the noise, everyone. Closing as invalid -- I was really running out of memory. /me going to the corner