I can confirm this bug. I also had this problem once, and I'm sure
it's in the new code. I'm pretty sure I can still fix this for the
stretch release.
--
regards Thomas
> There was a problem when upgrading from 6.20.00-5 (sorry for the German):
This is a problem in the postrm of 6.20.00-4 which is fixed in
6.20.00-5. Since I cannot fix 6.20.00-4, you should just ignore this.
--
regards Thomas
Package: am-utils
Severity: wishlist
There's the new upstream version 6.2 from October 30, 2014 which now
supports NFS v4.
BTW, the current version in unstable is 1065 days old, and has not yet
entered testing. Is this package orphaned?
--
regards Thomas
> I forgot to note that the aufs-module can be found in the package
> aufs-dkms (source: aufs) in Debian stretch+.
I thought aufs was replaced by overlayfs. But you are right, there's
still aufs in the dkms package.
--
regards Thomas
Hi Edgar,
thanks for your patch, but it does not apply cleanly to the master
branch of FAI. Could you please adjust your patch?
Here's the output when applying the patch using the patch util:
patching file fcopy
Hunk #2 succeeded at 309 with fuzz 2 (offset 10 lines).
Hunk #3 FAILED at 324.
1
This new behaviour will break some FAI installations. It would be very
good, if such an APT::Get:: option will be added. I'm hoping to have
such an option also for apt-get since a long time. This option should
also ignore package names which are unknown, for example due to a
typo. Using the new
> On Thu, 25 Aug 2016 01:59:47 +0200, Axel Beckert said:
> So let's try to define the exact behaviour needed for automated setups
> where package lists might be not 100% fit what's available.
> * Installation requests for packages which don't exist in the
Just some addition info. This raid is created with metadata=imsm. This
is Intel Matrix Storage Manager metadata format, which creates a
container. Containers are not yet supported by setup-storage.
Do you really want to use this fake-hw-raid, or would it be OK to
force software raid (means
> On Thu, 25 Aug 2016 10:41:57 +, LAUTERWASSER Stefan
> said:
> I don't want, I have to. :-/
OK.
Before adding your patch, I will test it in my environment.
> Please let me know which kind and way you prefer for communication - do
you
> On Fri, 26 Aug 2016 11:45:16 +0200, deb...@jack.fr.eu.org said:
> There it is
> [1.926556] dracut:
> //lib/dracut/hooks/mount/99-mount-root.sh@29(mount_root): mount -t ext4
> -o rw,relatime,errors=remount-ro,data=ordered,ro,ro
> /dev/mapper/data-rootfs /sysroot
>
Hi Stefan,
I've tested your patch with a raid10 configuration. No problems so
far. I had to adjust your patch because it did not apply cleanly to
the master branch. Here's the commit
https://github.com/faiproject/fai/commit/31a7904235c69455b7c7603b521d23f8703e3b0b
I have one question. Is it
This fix will close the bug. It will be available in FAI 5.2.
https://github.com/faiproject/fai/commit/92e0dae8cc09088fa06522c217104cb1f2c53d5a
--
regards Thomas
IMO the justification on the severity is strange. A package should not
need to handle every case of corrupted parts on disk or otherwise
strange broken things. In this case your whole system is broken, and
you should fix this anyhow, but not blame netfilter-persistent for not
handling this
The problem is, that when using a qcow2 image, the OS reports the disk
size to be like 192k, not 10G. This also happens on Debian jessie and
using different partitioning tools like parted, cfdisk or looking at
/proc/partitions. Therefore it's not the fault of setup-storage or
parted.
It can be
You can use parted to see if setup-storage has created the partitions
with optimal alignment. It works like this:
(parted) align-check optimal 1
1 aligned
--
regards Thomas
> On Fri, 30 Sep 2016 14:19:49 +0100, Alastair McKinstry
> said:
> I have patched 2 RC bugs that are threatening to get tcsh (and hence my
> packages) kicked out of testing.
> Please review the patch:
Sorry, I cannot review the patch, since I'm
> On Sat, 17 Sep 2016 17:13:28 +0200, "Manuel A. Fernandez Montecelo"
> said:
> The common case when things don't fail should be fast and more "correct"
> when informing about problems.
Yes, but this would still be only a workaround and no clean fix.
> On Thu, 25 Aug 2016 23:28:18 +0200, deb...@jack.fr.eu.org said:
> My rootfs lives on a logical volume, uncrypted
Strange. Doe you have any error messages?
You can debug the boot process by adding rd.debug to the
kernel command line. Then you will get a lot of output. The log is
written
> On Sun, 21 Aug 2016 23:18:17 +0200, Jack said:
> Without dracut-config-generic (so, with hostonly=yes), it fails to boot
on my ext4 rootfs, as the kernel module crc32c was missing
I cannot confirm this bug. On Debian testing this works for me without
any
> On Mon, 08 Aug 2016 14:33:52 +0200, Frederik Himpe
> said:
> I installed fai from backports on my Debian Jessie system. When running
fai-make-nfsroot, now it fails with:
> E: Unable to locate package dracut-config-generic
I've just uploaded a new dracut
> On Sun, 25 Sep 2016 01:24:59 +0200, Nicolas Braud-Santoni
> said:
> I was working on this today at the Salzburg BSP and should finish
> tomorrow.
Any progress so far?
--
regards Thomas
The host www.aeolusproject.org does not have an entry in DNS any more.
It seems that the source is now hosted at
https://github.com/redhat-imaging/imagefactory
--
regards Thomas
I've only moved the mount2dir executable from the fai-nfsroot package
to the fai-client package. The fai-client does not replace
fai-nfsroot.
To fix this bug is it only possible by defining a breaks or replaces?
Which breaks do I have to define then?
BTW, fai-nfsroot is never installed on a
> On Mon, 07 Nov 2016 17:36:41 -0500, Sam Hartman
> said:
> Currently, the sample configuration namespace has a shell script to
> restore the common capabilities found in base files; see
> scripts/DEBIAN/20-capabilities.
In this script, I'm doing the
Another way wouldbe to copy /etc/fai/NFSROOT into the CS. This should
also add the packages defined in this config file to the partitial mirror.
--
regards Thomas
Package: ca-certificates
Version: 20161102
My /usr/local file system is mounted read-only via NFS. This results
in an error:
stretch[~]# dpkg --configure ca-certificates
Setting up ca-certificates (20161102) ...
chmod: changing permissions of '/usr/local/share/ca-certificates': Read-only
file
I found the thread on the linux-fai mailing list and also the code
that added efi support into setup-storage. In the end we remove the
code from FAI, since it was not needed any more.
It's much easier to get a partition of type ESP. Just create a disk
gpt disk label, a partition of type vfat that
>>>>> On Tue, 08 Nov 2016 07:47:22 -0500, Sam Hartman <hartm...@debian.org>
>>>>> said:
>>>>> "Thomas" == Thomas Lange <la...@informatik.uni-koeln.de> writes:
Thomas> Just as a short note. There's the commands fai-deps(
Just as a short note. There's the commands fai-deps(8) which can be
used to define dependencies inside classes. It's available in FAI but
not used (means called) by default.
--
regards Thomas
.
-
Fully Automatic Installation - FAI
5.2 (c) 1999-2016
Thomas Lange <la...@informatik.uni-koeln.de>
-
Starting FAI execution - 20161103_
In FAI 5.3 I've add a test, to prevent fai-diskimage to run if a LVM
partition is detected on the host system.
https://github.com/faiproject/fai/commit/dce458cfc7c6b3bf20b5f235f08f8c68fd875644
--
regards Thomas
This bug seems to be related to #819725.
It seems that I used an older version of mkfs when creating the root
fs, so this error did not appear in my enviroment.
I will dig deeper into it to see how to fix this.
--
regards Thomas
I wonder why you need to source /usr/lib/fai/subroutines for importing
the ifclass subroutine. If your scripts are bash scripts, this
function should be already available.
--
regards Thomas
> On Sun, 08 Jan 2017 09:03:54 +0100, intrigeri said:
> I can reproduce neither the original issue, nor the LUKS passphrase
> one anymore.
Hi,
does the same apply to #797315, the molly-guard case?
--
regards Thomas
Package: fai-server
Version: 5.3
Severity: minor
When calling fai-diskimage, FAI sets NIC1 to the nic name of the host
system, which maybe wlan0 or eth4 or some other value. NIC1 is used
for creating /etc/network/interfaces.
This can easily be fixed by setting NIC1=eth0 in a class/*.var file.
--
I wonder why rinse does not use ipv6. rinse uses wget for getting the
files. Can you try to add --prefer-family=IPv6 to the system call
inside rinse. Or maybe you have a ~/.wgetrc which is not used when
called by rinse.
--
regards Thomas
There's still no fix. Do you need help for a fix?
--
regards Thomas
Since we need to build pdf, I will use asciidoc-dblatex as build
dependency.
--
regards Thomas
Hi Michael,
the status is unchanged. Since I never did anything using triggers, I
do not feel competent writing it for dracut. So, this is a request for
help. And I'm happy to include a patch. Maybe Ben has an idea how to
do this properly.
--
regards Thomas
Package: thunderbird
Version: 1:45.8.0-2
Severity: important
Tags: patch
* What led up to the situation?
I upgraded stretch and got thunderbird instead of icedove.
* What exactly did you do (or not do) that was effective (or
ineffective)?
I launched thunderbird.
* What was the
+
+ * control: fix Recommends, Closes: #852083
+
+ -- Thomas Lange <la...@debian.org> Wed, 22 Mar 2017 18:52:25 +
+
dracut (044+241-2) unstable; urgency=low
* add patch for dkms aufs modules, Closes: #855370
diff -Nru dracut-044+241/debian/control dracut-044+241/debian/control
---
Today, rpc.idmapd is not strictly needed for nfsv4.
--
regards Thomas
ile: make sure root directory has mode 755, Closes: #854654
[ Thomas Lange ]
* install_packages: add handling of packagename:arch, Closes: #855799
* control: fcopy needs libfile-lchown-perl
* Makefile: fix build error because of missing lchown.pm
diff -Nru fai-5.3.4/bin/fai-make-nfsroot
Package: fai-client
Version: 5.3.5
Severity: important
Tags: patch
Softupdates do not work in fai 5.3.5. You get an error message during
task_instsoft, "Undefined subroutine ::insert called at
/usr/sbin/install_packages line 547.". If you edit that file and
change the subroutine call from insert
00 +0100
+++ fai-5.3.6/debian/changelog 2017-04-20 12:40:03.0 +0200
@@ -1,3 +1,17 @@
+fai (5.3.6) unstable; urgency=low
+
+ [ Thomas Lange ]
+ * install_packages: fix subroutine name, Closes: #860783
+thanks to John Heim for the bug report and fix
+
+ [ Ian Kelling ]
+ * fai
The fix works for me. Please try to fix it soon, so the fixed version
will be going into the stretch release.
--
regards Thomas
This was fixed in upstream and I want to point to the commit and bug
report upstream:
https://bugzilla.redhat.com/show_bug.cgi?id=1190098
ba665de454e7d6ef1da4ba95308fb9f65c7a9497
--
regards Thomas
Package: debootstrap
Severity: normal
Tags: patch
I've wrote some patches, which speed up deboostrap a lot.
Changes are only made to /usr/share/debootstrap/functions, mostly in
the perl part, but also the number of lines that are proceeded by
perl are reduced via egrep.
Here are some time
I wonder why the sync was needed. I cannot find any info in the git
logs. It seems to be there for a long time. IMO this should be
removed, if noone can argue why it is needed.
--
regards Thomas
I'm not sure if he really wants to set the mtu for the DHCP or in
general for the interface. I do this in /etc/network/interface using
this line
pre-up /sbin/ifconfig eth0 mtu 9000
IMO you can close this bug.
--
regards Thomas
I will be working on this new version during DebCamp in Montreal
before DebConf. This will be in the beginning of august.
--
regards Thomas
I have patched fetch-basefile so this should now work as
expected. Instead of an eval and indirect shell variables, I just use
an array and loop over this. No performce penalty unless you have some
thousands of classes defined for a client and a million of base files ;-)
Here are the diffs:
Package: bash
Version: 4.2+dfsg-0.1+deb7u4
Testing this on Debian wheezy with
GNU bash, version 4.2.37(1)-release (x86_64-pc-linux-gnu)
host[~]$ BAR=foo
host[~]$ BAZ=${BAR:1:1}
host[~]$ echo $BAZ
o
This seems to work without any problems. I;ve tested the same on
Debian jessie and stretch.
> On Thu, 13 Jul 2017 17:41:56 -0400, Arcady Genkin
> said:
> guessing that the hyphen in the host name is causing the problem (the
hostname
> is "eddie-vm.teach.cs.toronto.edu" which defines a FAI class of the same
name).
Hi Arcady,
the problem is, that
> On Wed, 26 Jul 2017 11:45:20 -0400, Arcady Genkin
> said:
> I think that the
> approach to replace hyphens and any other illegal characters with
underscores
> is a very good one.
And we have dots in FQDN, which are currently also not allowed in
class
Doing gpg --import ~/.gnupg/secring.gpg works for me perfectly.
Maybe we can close this bug now.
--
regards Thomas
Ok, I've applied it, but it's not yet pushed.
Thanks for the explanation and quick feedback.
--
regards Thomas
I've managed to create an image (using the FAI 5.4~beta2 version) with
following change:
--- Volumes.pm 2017-01-04 21:18:01.351924832 +0100
+++ /usr/share/fai/setup-storage/Volumes.pm 2017-08-18 22:51:39.620984379
+0200
@@ -401,6 +401,7 @@
use Linux::LVM;
use Cwd qw(abs_path);
This problem is described in https://github.com/dracutdevs/dracut/issues/118
and is fixed upstream in commit f0bfada399c2c3269de353be5af37224349839a0
IIRC this can be ignored without causing any problems.
--
regards Thomas
It would be very nice if you could upload a new version fixing this
issue in the next two weeks.
--
regards Thomas
Is there any reason why emacs25 is not yet an alternative in the
dependency list?
If not, I would raise the severity of this bug to important.
--
regards Thomas
The description says:
This is a metapackage that will always depend on the latest
recommended Emacs release.
This is not true any more.
--
regards Thomas
JFTR, this was added in commit 72b59db126ebf39.
Here's the comment in this commit:
Align partition sizes to sectors or
user-requested alignment on all disk labels, unless on DOS and preserving
some partitions (stay with cylinder-alignment in those cases).
Currently I fear to change
Here's the patch that automatically detects if grep supports the Perl
regex.
0008-automatic-detect-if-grep-support-perl-regex.patch
Description: Binary data
--
regards Thomas
My last patch was broken. Here's a new working version.
0008-automatic-detect-if-grep-support-perl-regex.patch
Description: Binary data
--
regards Thomas
I could also add a test, to check if grep supports -P. If not, don't
use it. Does a non-GNU grep use -P for other things?
--
regards Thomas
Thanks for reporting. I will change the severity to normal, because
it's not an important issue.
--
regards Thomas
I found another bug in the code. So this still needs more tests.
--
regards Thomas
fai 5.4~beta23 has now a fix. Instead of vgremove I do a vgchange -a n
$vg, which leaves the disk image bootable.
Changes:
https://github.com/faiproject/fai/commit/f577e19d90210cf1165c58870804c80f47d8212f
--
regards Thomas
Package: fai-client
Version: 5.4
Severity: normal
Tags: patch
when calling fai-diskimage and not using a base file debootstrap is
called with wrong parameters. From the logs:
Calling debootstrap sid http://deb.debian.org/debian/
/usr/lib/fai/subroutines: line 957: call_verbose: command not
Hello Benno,
thanks for your information and for the tip with the package manager and the
"nopause" option for nano-2.8.0. But the problem is not so bad for me, so I can
wait for the next Debian stable release, which contains a newer version of
nano. Sorry for the confusion: I am the author of
I've pointed upstream to this patch and asked if upstream likes to
include it or to make micorcode loading more generic.
--
regards Thomas
I've implemented a fix for that. Now fai-diskimage detects which
volume group your host is using and sets the variable SS_IGNORE_VG
with a list of vg's to ignore. This is then used by setup-storage.
It's fixed in FAI 5.4~beta9. You can try the new beta version
by adding this line to your
> On Thu, 7 Sep 2017 18:28:33 +0200, Tiago Ilieve
> said:
> Do you mind to share how you identified those slower parts of the
> code? Have you used a profiler or a similar tool?
I didn't used a profiler. I just look at the code and saw that the
IMO we can close this bug, because this is nowadays implemented
(pretty similar) by qemu-debootstrap.
--
regards Thomas
I forget something. During the run of debootstrap, I observed via
repeated ps -efl calls that often perl is running for a longer
time. Then I tried to use strace to see which exec(ve) calls takes a
long time. But I didn't managed to find out, instead I saw read calls
taking long. So I decided to
Here's another patch, that improves speed. It needs grep -P which is
not available in busybox. But I do not know if debootstrap is used in
a busybox enviroment.
The execution time is now
real0m39.141s
user0m38.924s
sys 0m2.980s
So the user time drops from 50 to under 40 seconds on
Package: debian-cloud
Tags: patch
In fai-cloud-image, there's a synatx error in a debconf preseeding
file. This may cause grub to hang during boot. Attached is the fix
0001-double-space-single-space.patch
Description: Binary data
--
regards Thomas
> On Sun, 03 Sep 2017 14:15:00 -0700, Noah Meyerhans
> said:
> it would be
> really helpful to have some kind of file templating system integrated into
> FAI, possibly in the fcopy program directly.
Any suggestions, which template system to use? Or which
Here's another patch which makes debootstrap a little bit faster.
0009-in-this-part-we-only-need-the-package-name-and-prior.patch
Description: Binary data
--
regards Thomas
Package: fai-server
Control: reassign 877988 dracut
Moving this bus to dracut, since it does not affect FAI. I'm not sure
if this causes problems in dracut at all.
--
regards Thomas
>>>>> On Mon, 27 Nov 2017 00:15:23 +0100, Thomas Lange
>>>>> <la...@informatik.uni-koeln.de> said:
> ls -l /dev/disk/by-label gives
> Debian\x209.2.1\x20amd64\x20n -> ../../sr0
> This is a kvm virtual machine booting from CD.
Boo
> On Mon, 27 Nov 2017 15:10:25 +, Chris Nix said:
> 1. I had a preinst in sources.list that used 'envsubst' to replace
strings like "${some_variable}" from the environment
> 2. After upgrade, source.list are copied with fcopy -S, which will
replace strings
The ISO images contain a directory .disk with some files.
# cat .disk/info
Debian GNU/Linux 9.2.1 "Stretch" - Official amd64 NETINST 20171013-13:07
Inside the d-i calling blkid /dev/sr0 outputs something like this:
/dev/sr0: UUID="2017-10-13-13-09-57-00" LABEL="Debian 9.2 amd n" TYPE="iso9660"
> On Sun, 26 Nov 2017 23:02:23 +0100, Cyril Brulebois
> said:
> I'm pretty sure I did mention I would like to avoid having to mount
> anything, so this is definitely a no-go.
I missed that, sorry.
ls -l /dev/disk/by-label gives
Debian\x209.2.1\x20amd64\x20n
After loading addition components, the CD is mounted and we can see
which device it is. I would guess that the additional components will
be loaded from the same device as we booted the installer.
Still have to check it when booting from USB.
--
regards Thomas
Here's a small improvement of the man page.
https://github.com/faiproject/fai/commit/0280a115519e28b10087d695671be4749ab202b5
--
regards Thomas
Package: fai-client
Version: 5.5
When calling fcopy with -S the preinst script is not executed any
more. In subroutines in task_repository() we now use fcopy -S for
sources.list. When people are using a preinst for sources.list, it
does not get executed. You should update your config space when
> On Wed, 15 Nov 2017 17:05:20 +0100, Bastian Blank said:
> Please always use --xattrs for base tar operations. This allows the
> ping e.g. capability setting to remain.
IIRC I didn't activated them always, because older versions of tar did
not support this option.
I can confirm this bug.
I would like to see some comment from the package maintainer.
Running cups client and cups server on Debian stretch
ii cups-client 2.2.1-8 amd64 Common UNIX
Printing System(tm) - client programs (Sy
ii cups-common 2.2.1-8
> On Mon, 21 May 2018 22:34:22 +0200, Vincenzo Demasi
> said:
> The emergency recovery shell report 'failed to mount /sysroot'.
> I suspect the issue is caused by the missing crc32c module. Indeed
Can you please check with lsinitrd if the crc32c module is
> On Tue, 22 May 2018 09:22:10 +0200, Vincenzo Demasi
> said:
> well, in attachment you can find the output of lsinitrd for the broken
image and
> the good one too, maybe you are interested to check differences between
the two.
I could not find
The problem could be that you are using ext3, not ext4. The file
/usr/lib/dracut/modules.d/99fs-lib/module-setup.sh
includes a line
xfs|btrfs|ext4)
where you could add ext3 and give it a try.
--
regards Thomas
> On Tue, 22 May 2018 08:34:18 +0200, Vincenzo Demasi
> said:
> > Can you please check with lsinitrd if the crc32c module is missing
and
> I've checked with lsinitrd and the module is not in the image, then I
checked
> the good one generated by
It seems that /dev/loop is and indicator if fai is called by
fai-diskimage.
This is some debug output when running fai-diskimage
= shell: GRUB_PC/10-setup =
ainsl: appending to /tmp/fai-diskimage.BARhAu/etc/default/grub:
GRUB_DISABLE_OS_PROBER=true
++ chroot
Package: fai-doc
Version: 5.2
In GRUB_PC/10-setup and GRUB_EFI/10-setup we match $GROOT against the
string "hostdisk" which should detect if we are using FAI when creating
a disk image using fai-diskimage.
This does not work correctly, since grub-probe also outputs
hostdisk. if no
JFTR
https://github.com/faiproject/fai/commit/67fa14fdeb5d0453bd9fac433d031d55a66b5107
--
regards Thomas
Oh, we already had the option -N to disable the checking and removing
of unknown packages in install_packages. I've now added the variable
FAI_DISABLE_PACKAGE_NAME_CHECK which does the same as -N. You can set
this variable in class/*.var.
--
regards Thomas
Are apt-get and apt are also affected by the different behaviour when
using -t release or package/release?
You may want to ask the apt maintainers if they have a solution for
using a release name without using -t. Or to force a better
dependency resolution when a packages uses /release.
--
> On Fri, 26 Jan 2018 08:01:22 + (UTC), Jens Sadowski
> said:
> Description: tcsh crashes with a core dump when entering non-ascii
characters on the command line, like, e.g., "§". This happens with the tcsh
from the most up-to-date ubuntu package as
701 - 800 of 1142 matches
Mail list logo