IRC channel logs
2026-01-29.log
back to list of logs
<cosine>same behavior with savannah/master - boots on one vCPU; with two vCPUs, last message is `ahcisata0: 64-bit DMA` <cosine>damo22: that's what expected, right? <damo22>you can type trace/tu to see what happened <damo22>cosine: you missed my messages: before 64-bit DMA is displayed, it actually drops to db> shell but because its multicore, it continues and you dont see the shell but you can run trace/tu <cosine>damo22: I'm not used to IRC. The web client did log me out, but I went back over the logs and saw your response. <damo22>now we have a baseline to try to fix these failing tests <damo22>so i found a parameter that i can pass to javac that allows it to resolve symbols via compiling a missing class, so in theory if i pass the entire jdk java source tree to every compile step and reuse the class files, it should compile everything and resolve all symbols itself. However, there are STILL missing symbols, i think because some of the code is C++ and added "native" to java <damo22>hmm if im missing 1.5.0 class files for jdk5 maybe i can just borrow them from a binary release <damo22>to obtain enough to bootstrap the rest of jdk6 <gnu_srs1>youpi: No making /run being mounted by tmpfs did not change anything. But /dev/shm (/run/shm) is still mounted by tmpfs. How do I disable it for testing? <gnu_srs1>youpi: You can restart jobs on mahler by now. I have had some intermittent router problems on the host recently. <youpi>gnu_srs1: in /dev/default you can set RAMSHM to no <youpi>gnu_srs1: thanks for restoring access to mahler! <rrq>on the first start of hurd by crosshurd, I get a shell for running "native-install"... could I make that be automatic? <rrq>i.e. I can script the preparation of the disk image for qemu, and would now want it to automatically run that ./natiive-install on the first boot-up <rrq>it appears something runs /bib/sh (?) but I'm unsure how to change that into running a script <rrq>hmmm is it maybe using launch_something ... so it tries /runsystem first ? <rrq>maybe /usr/libexec/runsystem (?) <rrq>no that's a symlink to nowhere <rrq>one of /usr/libexec/runsystem.{hurd,gnu} perhaps <youpi>rrq: the long-term goal is to directly write the translator xattrs, mmdebstrap does this, making crosshurd just invoke mmdebstrap is on the todo-list, contribution welcome, there are details on the mailing list archive <rrq>ok. that would avoid the native-instal step I guess. <youpi>yes, everything is then doable from the linux host <rrq>I'll see if I can do something useful there... I still haven't had any luck how to install grub on the image file though; I resorted to extlinux with mboot <jab>damo22: maybe this is a dumb idea, but could you install guix on Debian Hurd. Guix has packaged openjdk, but I don't know if they ported to the Hurd... <gnu_srs1>Hi again: How do I disable /dev/shm to be mounted: cat /proc/mounts; <gnu_srs1>none /dev/shm /hurd/tmpfs writable,relatime,no-inherit-dir-group,no-sync,size=1545880K 0 0 <gnu_srs1>Setting RAMSHM=no in /etc/default/tmpfs did not help. An umount /dev/shm seems to not help, being a passive translator? <youpi>it's normally not a passive translator <gnu_srs1>showtrans /dev/shm;/hurd/tmpfs --mode=1777 50% <gnu_srs1>yay: Removing the /dev/shm translator made glxgears work in X windows!!! :) <gnu_srs1>Some output: wrlib: could not allocate shared memory segment, shmget: Permission denied <youpi> /dev/shm is normally with perms 1777 <gnu_srs1>How can I disable /dev/shm to be mounted on tmpfs for now? <youpi>now that you have removed the passive trnaslator and set RAMSHM=no, it's really not supposed to be mounted as tmpfs <sobkas> 475 - RunCMake.ctest_test (Failed) <sobkas> 538 - RunCMake.CTestTimeout (Failed) <sobkas> 539 - RunCMake.CTestTimeoutAfterMatch (Failed) <sobkas> 582 - RunCMake.CTestCommandLine (Failed)