Re: Build of devel/eclipse hangs
On Wed, Oct 05, 2005 at 04:43:07PM -0400, Kurt Miller wrote: > > I'll retry with a snapshot in a few days if nobody can reproduce > > this with -current. > > So far I can't reproduce it. I'm still working on ld.so. When that > stablizes I can come back to this. FYI: After having no luck with a snapshot some weeks ago (oct, 7th, IIRC), the port now builds fine with a recent -current. Just in case all this was an ld.so issue: $ cd /usr/src/libexec/ld.so/CVS && sed /^D/d Entries /dir.h/1.2/Mon Jun 2 19:38:24 2003// /dl_printf.c/1.13/Sun Jul 6 20:03:57 2003// /sod.c/1.22/Sun Oct 17 09:30:05 2004// /sod.h/1.1/Fri Jul 12 20:18:30 2002// /strtol.c/1.1/Sun Jul 6 20:03:58 2003// /util.c/1.18/Tue Aug 17 19:17:01 2004// /util.h/1.19/Sun Oct 17 09:30:05 2004// /dir.c/1.10/Fri Jun 17 22:21:56 2005// /ld.so.1/1.8/Tue Aug 16 21:15:03 2005// /Makefile/1.30/Sat Oct 1 10:18:43 2005// /library.c/1.51/Fri Oct 14 18:16:09 2005// /library_mquery.c/1.31/Fri Oct 14 18:16:09 2005// /dlfcn.c/1.70/Tue Oct 18 19:30:53 2005// /resolve.h/1.49/Tue Oct 18 19:30:53 2005// /library_subr.c/1.22/Sat Oct 22 14:10:41 2005// /loader.c/1.99/Sat Oct 22 14:10:41 2005// /resolve.c/1.45/Sat Oct 22 14:10:41 2005// Ciao, Kili -- What is this? Some kind of grep bitten by a radioactive spider? -- William S. Yerazunis about the crm114 interpreter in "The CRM114 Discriminator Revealed!"
Re: Build of devel/eclipse hangs
On Tuesday 04 October 2005 06:18 pm, Matthias Kilian wrote: > On Mon, Oct 03, 2005 at 05:06:20PM -0400, Kurt Miller wrote: > > In that case, I made some commits today that might effect this. Not > > sure if mirrors have them yet. make sure your tree has revision 1.59 or > > later of src/libexec/ld.so/dlfcn.c to be sure you have the latest > > corrections. > > No change, same problem as before. And, yes: I updated and rebuilt > all (kernel and complete user land). However, I didn't rebuild the > jdk, since I didn't see any changes in gcc and binutils for the > last three days. > > I'll retry with a snapshot in a few days if nobody can reproduce > this with -current. So far I can't reproduce it. I'm still working on ld.so. When that stablizes I can come back to this. -Kurt
Re: Build of devel/eclipse hangs
On Mon, Oct 03, 2005 at 05:06:20PM -0400, Kurt Miller wrote: > In that case, I made some commits today that might effect this. Not > sure if mirrors have them yet. make sure your tree has revision 1.59 or > later of src/libexec/ld.so/dlfcn.c to be sure you have the latest > corrections. No change, same problem as before. And, yes: I updated and rebuilt all (kernel and complete user land). However, I didn't rebuild the jdk, since I didn't see any changes in gcc and binutils for the last three days. I'll retry with a snapshot in a few days if nobody can reproduce this with -current. Ciao, Kili
Re: Build of devel/eclipse hangs
On Monday 03 October 2005 04:40 pm, Matthias Kilian wrote: > On Mon, Oct 03, 2005 at 04:09:08PM -0400, Kurt Miller wrote: > > Thanks for the detailed report. The problem you are seeing looks > > exactly like a problem that was fixed about 10 days ago. Are you > > sure you are running -current userland? > > Yes, I'm pretty sure, since I'd other trouble with qt's uic (which > seemed to be related to ld.so), so I rebuilt everything. In that case, I made some commits today that might effect this. Not sure if mirrors have them yet. make sure your tree has revision 1.59 or later of src/libexec/ld.so/dlfcn.c to be sure you have the latest corrections. > However, I've a make build running right now and will retry to build > eclipse after this. > > NB: I forgot to mention that I prefer FLAVOR=native_bootstrap when > rebuilding the JDK. > > Ciao, > Kili
Re: Build of devel/eclipse hangs
On Mon, Oct 03, 2005 at 04:09:08PM -0400, Kurt Miller wrote: > Thanks for the detailed report. The problem you are seeing looks > exactly like a problem that was fixed about 10 days ago. Are you > sure you are running -current userland? Yes, I'm pretty sure, since I'd other trouble with qt's uic (which seemed to be related to ld.so), so I rebuilt everything. However, I've a make build running right now and will retry to build eclipse after this. NB: I forgot to mention that I prefer FLAVOR=native_bootstrap when rebuilding the JDK. Ciao, Kili
Re: Build of devel/eclipse hangs
Hi Matthias, Thanks for the detailed report. The problem you are seeing looks exactly like a problem that was fixed about 10 days ago. Are you sure you are running -current userland? -Kurt On Monday 03 October 2005 03:00 pm, Matthias Kilian wrote: > Hi, > > when trying to build deve/eclipse, the build hangs after some time. From > eclipse-sdk-3.1p2.log: > > [...] > > | [echo] TARGET: compiler > > [tons of verbose ant blurb] > > | [javac] [total 10074ms] > | [echo] UPDATE ecj.jar > | > | BUILD SUCCESSFUL > | Total time: 15 seconds > | Assuming RHEL CLASSPATH compatible. > | [echo] TARGET: compiler2 > | [echo] compilerArg -encoding ISO-8859-1 > | [echo] build compiler org.eclipse.jdt.core.JDTCompilerAdapter > | [echo] UPDATE ecj.jar > | > | BUILD SUCCESSFUL > | Total time: 13 seconds > | Assuming RHEL CLASSPATH compatible. > | [echo] Deleting jars to recompile... > > At this point, the java process just eats up CPU time. A quick > > ktrace(1)/kdump(1) gives the following: > |478 java EMUL "native" > |478 java CALL sigreturn(0xcfbd12ec) > |478 java RET sigreturn JUSTRETURN > |478 java PSIG SIGSEGV caught handler=0x8c54e28 mask=0x0 > | addr=0x8293c000 trapno=2 478 java CALL write(0x4,0xcfbd12c7,0x1) > |478 java RET write -1 errno 35 Resource temporarily unavailable > |478 java CALL sigreturn(0xcfbd12ec) > |478 java RET sigreturn JUSTRETURN > |478 java PSIG SIGSEGV caught handler=0x8c54e28 mask=0x0 > | addr=0x8293c000 trapno=2 > > [...] > > And so on, ad infinitum. > > I'm using a two days old -current (oct. 1th, 15:00 UTC) on i386 and > the following limits: > > time(cpu-seconds)unlimited > file(blocks) unlimited > coredump(blocks) unlimited > data(kbytes) 716800 > stack(kbytes)32768 > lockedmem(kbytes)316673 > memory(kbytes) 948592 > nofiles(descriptors) 128 > processes128 > > malloc.conf is AJFG, USE_SYSTRACE is Yes, but I had the same problem > some days ago an tried with systrace and malloc.conf disabled, > without any success. > > All packages (except eclipse, of course) are uptodate, including > the 1.4 JDK. > > Could anyone please try to reproduce this problem? > > Ciao, > Kili
Build of devel/eclipse hangs
Hi, when trying to build deve/eclipse, the build hangs after some time. From eclipse-sdk-3.1p2.log: [...] | [echo] TARGET: compiler [tons of verbose ant blurb] | [javac] [total 10074ms] | [echo] UPDATE ecj.jar | | BUILD SUCCESSFUL | Total time: 15 seconds | Assuming RHEL CLASSPATH compatible. | [echo] TARGET: compiler2 | [echo] compilerArg -encoding ISO-8859-1 | [echo] build compiler org.eclipse.jdt.core.JDTCompilerAdapter | [echo] UPDATE ecj.jar | | BUILD SUCCESSFUL | Total time: 13 seconds | Assuming RHEL CLASSPATH compatible. | [echo] Deleting jars to recompile... At this point, the java process just eats up CPU time. A quick ktrace(1)/kdump(1) gives the following: |478 java EMUL "native" |478 java CALL sigreturn(0xcfbd12ec) |478 java RET sigreturn JUSTRETURN |478 java PSIG SIGSEGV caught handler=0x8c54e28 mask=0x0 addr=0x8293c000 trapno=2 |478 java CALL write(0x4,0xcfbd12c7,0x1) |478 java RET write -1 errno 35 Resource temporarily unavailable |478 java CALL sigreturn(0xcfbd12ec) |478 java RET sigreturn JUSTRETURN |478 java PSIG SIGSEGV caught handler=0x8c54e28 mask=0x0 addr=0x8293c000 trapno=2 [...] And so on, ad infinitum. I'm using a two days old -current (oct. 1th, 15:00 UTC) on i386 and the following limits: time(cpu-seconds)unlimited file(blocks) unlimited coredump(blocks) unlimited data(kbytes) 716800 stack(kbytes)32768 lockedmem(kbytes)316673 memory(kbytes) 948592 nofiles(descriptors) 128 processes128 malloc.conf is AJFG, USE_SYSTRACE is Yes, but I had the same problem some days ago an tried with systrace and malloc.conf disabled, without any success. All packages (except eclipse, of course) are uptodate, including the 1.4 JDK. Could anyone please try to reproduce this problem? Ciao, Kili