| Summary: | dev-java/icedtea-7.2.6.3-r2 hangs during awt build | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Jeroen Roovers (RETIRED) <jer> |
| Component: | [OLD] Java | Assignee: | Andrew John Hughes <gnu_andrew> |
| Status: | RESOLVED TEST-REQUEST | ||
| Severity: | normal | CC: | chewi, java, proxy-maint |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: | dev-java:icedtea-7.2.6.3-r2:20151220-160548.log.xz | ||
|
Description
Jeroen Roovers (RETIRED)
2015-12-21 07:59:40 UTC
I don't have physical access to timberdoodle so I always build without DISPLAY. I am familiar with this precise point in the build because I also had problems with it hanging. After much trial and error, this turned out to be because the kernel had CONFIG_COMPAT_BRK enabled. "On non-ancient distros (post-2000 ones) N is usually a safe choice." I was already very surprised that someone had enabled this so I'd be even more surprised if you have too. You can turn the option on or off without rebooting by setting /proc/sys/kernel/randomize_va_space to 1 (on) or 2 (off). Please test and report back. I have no idea what you are talking about. DISPLAY isn't set at all (and shouldn't be in a headless setup). CONFIG_COMPAT_BRK is disabled in the kernel. /proc/sys/kernel/randomize_va_space happens to be set to 2. None of these should matter. (In reply to Jeroen Roovers from comment #2) > DISPLAY isn't set at all (and shouldn't be in a headless setup). Right and I was implying that it's always worked fine for me. I have built icedtea on timberdoodle without DISPLAY countless times and the result of that is used to generate icedtea-bin. > CONFIG_COMPAT_BRK is disabled in the kernel. > /proc/sys/kernel/randomize_va_space happens to be set to 2. > > None of these should matter. CONFIG_COMPAT_BRK was definitely the culprit in my case so I thought I should mention it but evidently it's something else here. Were you suggesting that it does work if you do have a DISPLAY available? Probably not. When this happened to me, I was able to reduce the problem down to a very simple test program so give this a try. import javax.imageio.*; public class Test { public static void main(String[] args) { ImageIO.scanForPlugins(); } } After the build fails, compile that with /var/tmp/portage/dev-java/icedtea-7.2.6.3-r2/work/icedtea-2.6.3/bootstrap/jdk1.6.0/bin/javac and then run it with /var/tmp/portage/dev-java/icedtea-7.2.6.3-r2/work/icedtea-2.6.3/bootstrap/jdk1.6.0/bin/java. Note my original issue was reported upstream at http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=2693. Gah, forgot what I was going to say because I was interrupted a dozen times. I notice that you have selected HotSpot (implicitly by not enabling the cacao flag). As mentioned elsewhere, I have had trouble with HotSpot on timberdoodle even though upstream hasn't been able to reproduce the problem at all on RHEL. It is definitely a race condition because it works fine under "taskset -c 0" so you could try this or try enabling the cacao flag to see if it makes any difference. I have tried very hard to identify the cause but my best guess is a bad/old kernel and I'm not allowed to mess with that. ...and it exhibits itself as a sudden crash or segfault as soon as it tries to execute the newly-built VM. This issue is reported here: http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=2368 Sorry for the noise, my brain also suffers from race conditions. :) portage 13881 99.9 0.3 2158680 24948 pts/4 RNl+ 11:22 284:14 /var/tmp/portage/dev-java/icedtea-7.2.6.3-r2/work/icedtea-2.6.3/bootstrap/jdk1.6.0/bin/java -XX:-PrintVMOptions -XX:+UnlockDiagnosticVMOptions -XX:-LogVMOutput -Xmx1792m -Xms1792m -XX:PermSize=32m -XX:MaxPermSize=160m -cp /var/tmp/portage/dev-java/icedtea-7.2.6.3-r2/work/icedtea-2.6.3/openjdk.build/tmp/sun/sun.awt/awt/classes -Djava.awt.headless=true sun.awt.ToBin Maybe that isn't caused by a display or lack thereof since it should have access now and yet it's been churning for hours. (In reply to James Le Cuirot from comment #4) > Gah, forgot what I was going to say because I was interrupted a dozen times. > I notice that you have selected HotSpot (implicitly by not enabling the > cacao flag). I'll give that USE=cacao a go, then. I'm going to assume that icedtea-3 works now. icedtea-7 just never worked that well on ppc64 and we've since dropped the keyword. I guess this can be closed as 7.2.6.3-r2 is no longer in tree? (In reply to ernsteiswuerfel from comment #9) > I guess this can be closed as 7.2.6.3-r2 is no longer in tree? It is already RESOLVED. |
