IRC channel logs

2023-05-13.log

back to list of logs

<janneke>ACTION tries to upgrate the hurd sources to latest git in an attempt to sidestep the mach-defpager problem and stubbles upon the weirdest errors
<janneke>NULL not defined
<janneke>util/atoi.c:96:15: error: unknown type name ‘u_char’
<janneke>ACTION keeps patching for now, but whut?
<janneke>hmm, it's _POSIX_SOURCE that's not defined in include/sys/types.h
<janneke>and __KERNEL_STRICT_NAMES being defined in linux/dev/include/linux/types.h
<janneke>linux/src/drivers/net/de4x5.h:898: warning: "FALSE" redefined
<janneke>grmbl
<janneke>test -s gnumach-undef-bad: puts, putchar
<janneke>ACTION gives up for now
<biblio>luckyluke: I tested with kdb. In can see the fault is from acpi user space. I am not sure if it is before loading acpi ELF or inside acpi main function. I could not build acpi.static in current HURD code.
<biblio>luckyluke: https://paste.debian.net/1280160/ I think it is before loading acpi ELF.
<biblio>EX_WRONG_ARCH 6001 /* valid executable, but wrong arch. */
<biblio>correct error output https://paste.debian.net/1280161/ (please skip the prev one as gnumach.gz had --disable-user32)
<janneke>FWIW jpoiret civodul; this doesn't even build:
<janneke>4e497b00ad gnu: gnumach-headers: Update to 1.8-0-8cc2ed2eec.
<janneke>b17e122109 gnu: mig: Update to 1.8-0-g04bfe7a912.
<janneke>(updating to latest git)
<janneke>=> <janneke> test -s gnumach-undef-bad: puts, putchar, ... and more
<janneke>so jpoiret, quite an achievement you found a set of package versions that work...
<janneke>*that build
<jpoiret>janneke: yes, newer hurd requires a newer glibc, which requires newer mig. I was pretty lucky to find something that builds, but we don't really know if the set of versions I chose was ever supposed to run properly
<janneke>jpoiret: right
<janneke>i guess we shouldn't complain, it's prolly a sign of active development; the hurd<->glibc dependency is especially tricksy i guess
<janneke>jpoiret: so, i tried to bisect a bit, reverting your patches one by one, upgrading to newer versions, but am a bit at loss right now
<jpoiret>janneke: I'm currently trying to build a newer set of packages with glibc 2.37, and cleaning up some messy stuff like hardcoded --host=i586-pc-gnu arguments
<janneke>jpoiret: nice, the forward flight, i like it
<gnucode>since the hurd project no longer plans to have a 64 bit kernel with a 32 bit userspace...is the hurd project still interested in portable RPC interfaces?
<gnucode>Just out of curiosity what is the usefulness in having "mig portable rpc declarations"?
<gnucode>open_issues/mig_portable_rpc_declarations
<gnucode>hmmm looks like "package-list-packages" crashes emacs on the console on the Hurd.
<gnucode>or maybe not crashes it, but does some weird things so that I had to force kill emacs
<gnucode>but it looks like debian has magit packaged to no worries either way!
<biblio>does anyone know how to build acpi.static ?
<jpoiret>anyone would be willing to lend a hand to the Guix port of Hurd? After recently upgrading to glibc 2.35 and gcc 11, I had to resort to ugrading mig/gnumach/hurd to make it build, but now it's just not booting properly. It's hanging while trying to spawn mach-defpager, and I don't really know what's happening/why
<jpoiret>it seems it's just waiting on some msg, but I don't know how to debug further
<biblio>jpoiret: for gnumach you may try ../configure --enable-kdb (not sure it will help.)
<jpoiret>We have kdb enabled, that's how I'm seeing that every thread is just waiting on some message
<jpoiret>gnumach is just idling
<jpoiret>looking at the debian packaging there seems to be a couple of patches applied on top, but we mostly do the same boot-up process
<jpoiret>it may be that we forgot to start an essential server but I'm not seeing anything right (it doesn't help that I can't find any doc on what is needed for the pager)
<jpoiret>Being able to inspect the IPC from kdb would be a start i think, but I don't know how to do that
<biblio>jpoiret: could you please check if you have acpi.static present in your hurd build.
<jpoiret>we have a non-static acpi
<biblio>jpoiret: please try to install libacpica-dev and try to build hurd again. it should generate acpi.static as well
<biblio>jpoiret: make acpi (full hurd is not needed i think)
<jpoiret>biblio: i'm using Guix to build Hurd. Are you saying a static acpi is important here?
<jpoiret>I would've expected it to not really matter at this point
<biblio>jpoiret: I was able to boot with acpi and ext2fs modules. once I skipped acpi - i did not manage to get login prompt. It could be my local config issue.
<jpoiret>i will try, thanks for the pointer
<jpoiret>didn't really help after setting those translators
<jpoiret>well, just got it to boot, and I think that it comes from a very stupid misconfiguration on our part :)
<biblio>jpoiret: only with ext2fs ?
<jpoiret>I haven't tried to remove acpi from the default translators yet but I don't think that it was our original issue
<jpoiret>got to figure how to use dde now :)
<biblio>ok
<biblio>jpoiret: i was able to boot with ext2fs without acpi.