The build process quit with an error. It compiles successfully the vm, but it blocks at some point while recompiling the java library. I can't figure out what cause the problem, but the problem I have a dual-P3 450, so I ran emerge using MAKEOPTS="". I have 384MB of RAM, but just to be sure not running out of memory I stopped every other process (and only 1MB of swap got used). ... and yes, I unmerged lesstif and openmotif :) In fact I also linked /usr/bin/zip, /usr/bin/unzip and /usr/bin/cpio to /bin/zip, /bin/unzip and /bin/cpio, 'cause when sometime ago I build the jdk from scratch this step was definitely needed. For the same reason I also emerged tcsh. The problem is perfectly reproducible. It happened 4/4 times, exactly at the same point ... and about 3 hours after starting emerging! :((( I redirected the output to a file. It's very long (about 3MB) so I'm going to attach just the beginning (showing the environment variables) and the end (showing the error). If needed I can post the whole log.
Created attachment 5788 [details] log of the beginning of the build process
Created attachment 5790 [details] log of the (sad) ending of the build process
I am very sorry for you... The problem looks like a segfault in the compiled version of javac. This could be for a number of reasons. What is your toolchain like, glibc/gcc/binutils versions would be helpful. It seems that the java code reacted badly to one (or more) of the above packages and thus the resulting binary it compiled was a dud That is all I can think of, sorry, good bug report though, you have covered most of the info about your system.
Well, I tried to compile using glibc-2.2.5-r7, binutils-2.13.90.0.4, gcc-3.2-r2 (or maybe -r1). Yesterday I upgraded to gcc-3.2-r4, which come with some new patch ... so right now the jdk is compiling ... guess I'll have news (hope good) in couple hours ;)) Btw in July I got the jdk successfully compiled (from scratch) using gcc3.2pre (was a gentoo 1.3beta system). PS: thanks for the appreciation about my report ;) in fact I'm a computer engineer ... I know quite well what is like to fix bugs with dumb bug reports ;))
The java ebuild has a patch for glibc-2.3+ so that could be it, but reading the patch it looked glibc-2.2 save, I could be wrong ..... but glibc-2.2.5 could have bugs, there was the atexit bug. Well all I can say now is it works for ppl using the latest testing gentoo gcc and the redhat 8.0 gcc in CVS and binutils 2.13.90.0.10 and 2.13.90.0.14 all ontop of glibc-2.3.1 Thats about all I can think of, if it ever fails again you could try to go the the compile directory and try to resume it and/or gdb the errant javac command.
Aaaah .... shame on me ... ;) at last I solved it! It wasn't a matter of tool chain. I moved to 'unstable' and nothing changed. But using the new tool chain the ebuild failed giving a more explicative error: it was failing trying to get info about a process ... guess why? GRSECURITY!!! I removed it from the kernel et voila'!!! It was probably a problem related to restrictions accessing the /proc filesystem ;) Maybe it would be worth to show some kind of warning ... just to avoid users wasting sooo much time! ;)) Great ebuild though!!! Thanks!
Seems to be fixed, so closing this bug.
