IRC channel logs

2019-12-03.log

back to list of logs

***Server sets mode: +nt
***wleslie_ is now known as wleslie
<gnu_srs1>youpi: A hurd image booted nicely with the cross-built ext2fs.static. What's next?
<youpi>well, I don't know what your eventual goal is
*wleslie applause
<gnu_srs1>You proposed that I test the cross-built ext2fs.static on another box.
<gnu_srs1>The cross-built image does not boot, as reported before.
<gnu_srs1>Hangs executing the runsystem script.
<youpi>then as I said yesterday, debug the script
<youpi>most probably that's where the issue is, and not the fact that the binaries are cross-built
<youpi>possibly it just can't cope with missing translator entries in /servers or such
<damo22>youpi: i built libpciaccess0 0.16-1+hurd.1 without jlledom[m]'s cleanup patch, just the detect patch, and netdde crashed
<gnu_srs1>I wonder if I should compile hurd with or without utf8 support?
<gnu_srs1>Does it matter?
<damo22>SIGILL illegal instruction
<gnu_srs1>How do I create translator entries from Linux?
***inoaffep is now known as pinoaffe
<youpi>damo22: that's surprising, I don't remember joan's patch fixing such kind of thing
<youpi>gnu_srs1: I don't see why utf8 should matter at all
<gnu_srs1>damo22: I use libpciaccess-0.16 for cross-building. Is that version OK?
<youpi>gnu_srs1: as I said before, normally translator entries are now xattr
<damo22>gnu_srs1: depends which variant
<damo22>gnu_srs1: its currently in a bad statte
<gnu_srs1>I can cheat by creating them natively and copy them to the cross-built image.
<gnu_srs1>damo22: I use the upstream 0.16 version.
<damo22>upstream 0.16 seems to be fine for me
<damo22>but we are trying to clean it up and fix the detection mechanism
<gnu_srs1> https://xorg.freedesktop.org/releases/individual/lib/
<damo22>yeah that should work actually
<damo22>i have built that and it runs fine on my hurd system
<damo22>in debian
<damo22>but joan has done a lot of cleanup work and there is a new patch for detection, except it crashes netdde
<damo22>0.16-1+hurd.1 is currently broken
<youpi>what I don't understand is: it just works for me
<damo22>youpi: try ifdown /dev/eth0 and kill netdde, settrans -fg /servers/bus/pci and reset the translator on servers/bus/pci and ls it
<damo22>then bring up your interface again
<damo22>you should see it is busted
<damo22>make sure theres no stray libpciaccess symlinks or anything
<damo22>i had strange backup of libpciaccess.so.0.11.1 and it was pointing to the wrong version
<youpi>that just works here
<damo22>wow
<youpi>I'm using the e1000 qemu card model
<damo22>how many bytes is your /hurd/netdde
<damo22>mine is 3171212
<youpi>mine is 3158924
<damo22>uh oh
<youpi>version 0.0.20150828-8 of the netdde package
<damo22>im on 0.0.20150828-7
<youpi>I disabled the atp driver in -8
<youpi>iirc that was indeed required
<damo22>ok
<youpi>because it was trying to access i/o ports without asking
<damo22>hehe ok
<youpi>(and it's a parallel port-based network card, we really don't care )
<damo22>what, that thing must be older than me
<damo22>ok with that netdde 0.16-1+hurd.1 works with eth0
<damo22>ok with that netdde, libpciaccess 0.16-1+hurd.1 works with eth0
<damo22>but rump does not
<damo22>must be in pci-userspace repo
<youpi>possibly pci-userspace also accesses some i/o port without asking
<youpi>the difference with the detect patch is that it really uses the pci arbiter and not the x86 access method
<damo22>it says i/o space memory space enabled in rump verbose log
<youpi>otherwise if you run it as root it will suceed using the x86 access method, and thus enable i/o ports
<youpi>memory space is not a problem
<damo22>oh c080 is part of the io port range i excluded, if that helps
<damo22>which is the ahci controllers port
<damo22>maybe it still need that for probing?
<youpi>I don't see why it should
<youpi>it should be using libpciaccess for that too
<damo22>huh?
<youpi>that's a cofnig part
<damo22>it still needs root
<damo22>to get the privileged ports
<damo22>then it calls pci_system_init()
<damo22>afaik pci-userspace is supposed to implement a callback that asks for io ports
<damo22>so i am giving it access to all io ports except PCICFG io
<damo22>problem is, qemu uses port c080 for AHCI which overlaps with PCICFG2
<damo22>so i am not granting permission to rump to use that port
<damo22>i need to test if i give it access to PCICFG2 anyway, will it then probe the device correctly
<damo22>bed time
<youpi>damo22: but pci-userspace shouldn't have to tinker with these
<youpi>since it's the bios which will have configured the BARs
***coderobe8 is now known as coderobe
***Emulatorman_ is now known as Emulatorman
<gnu_srs1>Copying /servers/socket/1 from a hurd image does not work (as root): cp -p /servers/socket/1 .: cp: cannot open '/servers/socket/1' for reading: Permission denied
<gnu_srs1>How on earth can I create a passive translator, including /dev files?
<gnu_srs1>What does this mean? settrans: console: Device or resource busy
<gnu_srs1>and: libexec/console-run: /dev/console: Not a terminal
***rah_ is now known as rah