Package: openssh-server
Version: 1:8.9p1-2.1
Severity: important
Tags: patch
Dear Maintainer,
After accepting an ssh connection, the sshd process is killed and I see
the following in dmesg:
audit: type=1326 audit(1645794361.669:40): auid=0 uid=100 gid=65534 ses=1
subj==unconfined pid=8338
Package: fastboot
Version: 1:10.0.0+r36-1~stage1.3
Severity: grave
Justification: renders package unusable
Dear Maintainer,
$ fastboot devices
fake placeholder until fastboot builds
Closer inventigation shows that the fastboot binary has indeed been
replaced by a small shell script with zero
Package: libsbc1
Version: 1.3-1+b2
Followup-For: Bug #856487
Not a stack corruption.
This is miscompilation of sbc_analyze_4b_8s_armv6. gcc appears to look
into the asm function and decides that it does not clobber r3 (which the
normal ARM ABI says is call clobbered). The last out +=
Package: fail2ban
Followup-For: Bug #855920
Dear Maintainer,
Test failures occur if the build environment does not have the tzdata
package installed. The test harness hardcodes TZ="Europe/Zurich".
Patch below fixes this by encoding the equyivalent explicit timezone
rather than using the
Package: molly-guard
Version: 0.6.1
Severity: important
Tags: patch
Dear Maintainer,
Installation fails with:
trying to overwrite '/sbin/coldreboot', which is also in package
kexec-tools 1:2.0.9-1
Looks we're missingthe preinst/postrm hooks to setup the diversion.
-- System Information:
Oops, wrong patch (earlier, broken version). Correct one attached.diff -ur molly-guard-0.6.1/debian/molly-guard.postrm clean/debian/molly-guard.postrm
--- clean/debian/molly-guard.postrm 2015-08-29 18:37:14.0 +0100
+++ molly-guard-0.6.1/debian/molly-guard.postrm 2015-08-31
Package: dstat
Version: 0.7.2-4
Severity: normal
Tags: patch
Dear Maintainer,
I recently encountered a crash on a machine with a slightly wonky
terminal config.
If both termios.TIOCGWINSZ and curses.tigetnum fail in initterm(), then
we die in gettermsize() with
NameError: global name
I had the same issue, it was caused by the bump-mapped transparecy, which I
guess my graphics hardwre (Intel Atom/i915) can't handle.
Commenting out the background-bumpmap: line in
/usr/share/cinnamon/theme/cinnamon.css worked around the problem.
Paul
--
To UNSUBSCRIBE, email to
On 01/07/2011 01:26 PM, Paul Brook wrote:
Further investigation shows that it's actually the use of GUID partition
tables that cause confusion. grub-probe correctly identifies filesystem
and LVM device, but fails to map partitions for the constituent devices.
The lack of GUID partition
Further investigation shows that it's actually the use of GUID partition
tables that cause confusion. grub-probe correctly identifies filesystem and
LVM device, but fails to map partitions for the constituent devices.
The lack of GUID partition table support is not directly relevant to this
Package: grub-pc
Version: 1.98+20100804-11
Severity: important
Tags: patch
grub-mkconfig fails on a system where / can not be accessed by grub.
The is_path_readable_by_grub function only checks that the filesystem is
known to grub. In my case (lvm+raid) grub understands the filesystem, but
not
Package: perl
Version: 5.10.1-13
Severity: normal
Tags: patch
By default patchlevel.h includes git_version.h. The latter is generated by
configure.
However the debian/config.debian uses patchlevel.h before running configure.
This has always been fairly borken, but old GCC processed the rest of
Now it seems that the irrlicht library depends on those symbols
provided by libgcc_s.so.1 (and does not define them locally contrary to
what was seen by Aurélien in libvorbis in #462318) and of course
dpkg-shlibdeps complains because they can't be found in the symbols file.
...
So should I revert
Package: kformula
Version: 1:1.6.3-7+b1
Severity: important
kformula depends on kcontrol. kcontrol no longer exists in kde4,
rendering the package uninstallable.
-- System Information:
Debian Release: squeeze/sid
APT prefers unstable
APT policy: (500, 'unstable')
Architecture: amd64
Package: muscleframework
Version: 1.1.6-4
Severity: important
Tags: patch
Justification: fails to build from source
Build fails on armel because configure script does not recognize the
triplet arm-linux-gnueabi. Attached patch fixes this by relaxing the target
matching (accept linux-gnu* rather
Thanks a lot! Do the package work well when it has been built?
Yes. Both client and server seem to work as expected.
Paul
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Package: tightvnc
Version: 1.2.9-21
Severity: important
Tags: patch
Justification: fails to build from source
Package fails to build on ARM. Looks like confusion between
Arm32Architecture and ArmArchitecture in the configure scripts.
Patch below fixes the system detection file to generate the
Package: hal
Version: 0.5.10-5
Severity: important
Tags: patch
hald-addon-input does not release evdev device nodes when a device is
removed. Because hald still has the device open the kernel cannot free
the device node. After a number of plug/unplug cycles (typically 32)
runs out of device
On Thursday 27 December 2007, Daniel Burrows wrote:
Hm, I wonder if this works in all languages. Christian, do you have
any idea? Can we get away with ngettext() on the first item in a
comma-separated list?
My understanding is that fragf is printf-like, and the %F allows languages to
Package: aptitude
Version: 0.4.10-1
Severity: minor
Tags: patch
After doing an upgrade the following message was displayed:
There are now 1 update [-63].
Note incorrect use of the plural are. Attached patch fixes this.
I have double checked the following cases:
There is now 1 update [-63].
Package: kdm
Version: 4:3.5.8.dfsg.1-2
Severity: important
Tags: patch
When logging in on a remote xterminal via XDMCP kdm segfaults, causing the
session to die and the greeter to be redisplayed.
A backtrace of the crash (below) shows that we crash trying to create a
bogus DBUS entry. Further
Incremental apt updates use ed style diffs. i.e. the pdiff is a script
suitable for feeding into the ed line editor. Conventionally there is also
a restricted versions of ed called red which implements only a safe
subset of ed commands. Apt includes a partial (Reduced/even more Restricted?)
Package: nfs-kernel-server
Version: 1:1.1.1-1
Severity: important
Running exportfs -r segfaults if any warnings or errors are emitted.
For example, if /etc/exports contains:
/ 127.0.0.1
Normally this would display a warning about missing sync arguments. Now
it segfaults. Similar segfaults
Package: man-db
Version: 2.4.4-4
Severity: normal
Manpaths containing a plus character ('+') are broken. They are not
searched when looking for manpages.
To reproduce:
$ mkdir man++
$ cd man++
$ mkdir man1
$ cp /usr/share/man/man1/man.1.gz man1
$ man -M `pwd` man
No manual entry for man
See
Package: manpages-dev
Version: 2.62-1
Severity: normal
Tags: patch
The longjmp(3) function may cause unexpected things to happen to local
varables modified after the call to setjmp(). I've encountered this
issue several times, so it seems worthy of a mention in the manpage.
Attached patch adds
Package: apt
Version: 0.6.46.3
Severity: important
Tags: patch
The Building dependency tree... takes an extremely long time on
may arm machine. Firther investigation shows this is spending most of
its time in gettimeofday(), which is quite a slow function on this
machine.
apt calls
Package: gtk-gnutella
Version: 0.96.1svn12109-1
Severity: serious
Tags: patch
Justification: no longer builds from source
Build fails on ARM with STATIC_ASSERT failures. The ARM ABI padds
structudes to 4-byte boundaries by default. The attached patch adds
__attribute__((packed)) to inhibit
Package: gnome-presence-applet
Version: 0.3.1-1+b1
Followup-For: Bug #389292
Package also FTBFS on arm and i386.
The problem seems to be GalagoList.
Google suggests this should be defined in galago-list.h. However this
header does not exist in the libgalago-dev package.
-- System Information:
Package: linuxlogo
Version: 4.14-4
Severity: serious
Tags: patch
Justification: no longer builds from source
Build fails on arm with:
make[3]: Entering directory
`/build/buildd/linuxlogo-4.14/libsysinfo-0.1.0/Linux'
make[3]: *** No rule to make target `armv4l', needed by `cpuinfo.o'. Stop.
This problem seems to have resolved itself.
I don't know which change fixed it, but it works ok now.
Paul
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Package: gcc-4.1
Version: 4.1.1-16
Severity: important
Tags: patch
#pragma pack(1) doesn't work properly on ARM. eg.
#pragma pack(1)
struct (char c;} S;
sizeof(S) is 4, not 1.
This breaks schism (and possibly other packages).
http://buildd.debian.org/~jeroen/status/package.php?p=schisma=arm
Package: paintlib
Version: 2.6.2-7
Severity: important
Tags: patch
Justification: fails to build from source
Build on ARM fails with:
pliff85dec.cpp: In member function 'virtual void
PLIFF85Decoder::Open(PLDataSource*)':
pliff85dec.cpp:182: error: size of array is negative
The culprit is
Package: klic
Version: 3.003-gm1-4
Severity: serious
Tags: patch
Justification: no longer builds from source
On ARM the buildd is failing because it times out while running the
configure script. The patch below increases the timeout by a healthy
margin to avoid this.
Paul
---
This assumes that WORD is 2 bytes. This is not true on arm - it is 4
bytes. WORD needs to be defined to be 2 bytes, or uint_16 used instead
of WORD throughout.
Wrong. WORD is a typedef for uint16_t, which is 2 bytes.
This is a gcc bug, pragma pack(1) doesn't.
See #394703
Paul
--
To
Package: pnet
Version: 0.7.4-1
Severity: serious
Tags: patch
Justification: no longer builds from source
Build fails on ARM with
Error: Rn must not overlap other operands -- `swp r3,r4,[r3]'
The error is correct. This is coming from a broken inline assembly
statement. Patch below fixes it.
Package: mrxvt
Version: 0.5.2-1
Severity: serious
Tags: patch
Justification: no longer builds from source
Unlike many other targets, ARM does not allow $ symbols in identifiers.
The only use of these appears to be for some third parts syayic code
analysis system. The patch below disables it on
Package: sysklogd
Version: 1.4.1-19
Severity: important
Tags: patch
/etc/init.d/sysklogd reload
gives the following error:
start-stop-daemon: unrecognized option `--quiet--signal'
Try `start-stop-daemon --help' for more information.
This command is used by (and fails from) the daily sysklogd
Package: xserver-xorg-video-nv
Version: 1:1.0.1.5-2
Severity: normal
XV output is extremely slow when using the nv driver.
When playing a video with xine or vlc using xv output the Xorg
process uses a large amount of CPU. With videos larger than about
640x480 it hits 100% CPU (2GHz amd64) and
Package: debian-installer
Severity: important
The sparc boot image on the etch beta-3 businesscard and netinst CDs
does not contain the isofs module for 32-bit kernels.
This makes the cds effectively useless on 32-bit sparc machines as teh
installer can't mount the cd. The SCSI drivers load
Package: vde
Version: 1.5.11-1
Severity: grave
Justification: renders package unusable
Installation fails with:
Setting up vde (1.5.11-1) ...
useradd: invalid numeric argument 'vde-net'
adduser: `/usr/sbin/useradd -d /home/vde-net -g vde-net -s /bin/false -u
109 vde-net' returned error code 3.
Package: yaird
Version: 0.0.12-15
Severity: normal
When run inside a Xen DomU yaird fails with:
yaird error: unrecognised device: /sys/devices/xen
yaird error: unrecognised device: /sys/devices/xen/vbd-768
The patch below fixes this by ignoring xen devices. Debian xen kernels
build xen
Package: mdadm
Version: 2.5.2-7
Severity: normal
Tags: patch
checkarray --all gives the following error on a system with both RAID1
and RAID0 arrays:
checkarray: E: /sys/block/md4/md/sync_action not writeable.
Patch below fixes this by making --all only select devices that have
md/sync_action.
Package: manpages-dev
Version: 2.28-1
Severity: normal
The man page for ftruncate(2) says:
If the file previously was shorter, it is extended,
However this is not always true. The glibc documentation says:
The POSIX standard leaves it implementation defined what happens
Package: manpages-dev
Version: 2.19-1
Severity: normal
The sendmsg(2) and recvmsg(2) contain an incorrect desctiption of
struct msghdr.
The manpage says:
socklen_tmsg_controllen; /* ancillary data buffer len */
Whereas the actual implementaton (from {x86_64-linux-gnu/,}bits/socket.h)
The missing symbols are defined in libXxf86vm.a
kdegraphics certainly builds here, and the most recent upload built fine on
the buildds. We depend on libxxf86vm-dev, so everything should be
available. I'm not sure why you would get a different result on your
machine. Are you running X.Org
Package: kpovmodeler
Version: 4:3.5.1-2
Severity: serious
Tags: patch
Justification: no longer builds from source
Build fails on i386 with
g++ -Wno-long-long -Wundef -ansi -D_XOPEN_SOURCE=500 -D_BSD_SOURCE
-Wcast-align -Wconversion -Wchar-subscripts -Wall -W -Wpointer-arith
-DNDEBUG -DNO_DEBUG
Package: vim
Version: 1:6.4-006+2
Severity: minor
Tags: patch
The SVK version control system generates patches in a slightly different
format to CVS and SVN. Vim does not autodetect these as patch files.
A typical svk generated patch looks something like:
=== arm-dis.c
Package: ksirc
Version: 4:3.5.0-3
Severity: normal
Pressing Ctrl-V used to paste the contents of the clipboard into the
test entry box. Now it does nothing. Selecting Edit/Paste from the
menu works fine. Ctrl-X and Ctrl-C also work.
Ctrl-V works in other KDE applications (kmail, konqueror),
Package: bzip2
Version: 1.0.3-1.1
Followup-For: Bug #348266
Patch below fixes this for me.
Paul
diff -ur bzip2-1.0.3/debian/docbase bzip2/debian/docbase
--- bzip2-1.0.3/debian/docbase 2006-01-16 01:31:24.0 +
+++ bzip2/debian/docbase2006-01-16 01:27:00.0 +
@@
Package: knetload
Version: 2.3-3.1
Severity: important
Tags: patch
Some kernels use 64-bit counters for network traffic. This means
/sys/class/net/*/statistics/[rt]x_bytes can contain a number that
doesn't fit in an unsigned int.
The attached patch fixes this by reading the counter into a
Package: gdb
Version: 6.3.90.20051119-1
Severity: normal
After upgrading from libreadline5 5.0-11 to libreadline5 5.1-1 gdb
segfaults when the delete key is pressed. To reproduce load gdb and
press the delete key twice. Backtrace below
Paul
#0 0xb7f74f6d in _rl_dispatch_callback () from
Package: subversion
Version: 1.2.3a-1
Severity: normal
The combination of svn+ssh and automatic ssh connection caching is
broken.
In my ~/.ssh/config I have
ControlPath /tmp/.ssh-%h-%p-%r
ControlMaster auto
The first svn command succeeds, the second fails with
Control socket
Package: openssh-client
Version: 1:4.2p1-5
Severity: normal
Automatic connections caching breaks after ssh is terminated with
SIGKILL. This frequently happens when ssh is invoked as a subprocess of
some other (arguably buggy) application, eg. subversion (separate bug
already filed for that).
Package: kpdf
Version: 4:3.4.2-2
Severity: normal
When viewing large pdf files kpdf seems to use an unreasonably large
amount of X pixmap resources. After loading iand paging through a large file
(eg the IA32 SDM http://www.intel.com/design/pentium4/manuals/index_new.htm )
xrestop reports that
Package: ksirc
Version: 4:3.4.2-3
Severity: minor
Tags: patch
After joining channels /dishighlight is automatically switched off, i.e.
highlighting is enabled: *lol* becomes lol in bold. Although
/dishighlight mentions that the state will be saved.
Any chance of getting the fix for this
Package: nload
Version: 0.6.0-2
Tags: patch
Followup-For: Bug #300267
The attached patch should fix this bug. It makes nload use 64-bit
counters for total traffic, even on 32-bit hosts.
Per-interval measurememnts will probably still wrap at 2^32 but that's
harmless, and a much more invasive
Package: vim-full
Version: 1:6.3-086+1
Severity: normal
Tags: patch
The file /usr/share/applications/gvim.desktop launches gvim without -f.
This breaks using gvim as the default text editor, or via open with
from other kde applications. Typically these will write to a temporary
file, spawn
I took a quick look at your qvm86 source tarball (daily
snapshot from dad-answers).
It's currently intended to be bundled directly into qemu.
Is it possible to unbundle it somewhat?
The idea is:
- Only kqemu.h will be needed to build qemu with qvm86
support.
-
On Thursday 15 September 2005 09:46, you wrote:
tags +pending
thanks
On Thu, Jul 28, 2005 at 03:16:12AM +0100, Paul Brook wrote:
Autossh doesn't handle -- properly when parsing commandling options. It
checks for and eats the first --, but it still parses the subsequent
commandline
qvm86 support would be more appropriate. qvm86 is a GPL replacement for kqemu.
http://www.qvm86.org/
Paul
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Package: ksynaptics
Version: 0.2.0-1.1
Followup-For: Bug #301274
Better patch which keeps TouchPad::coastingSpeedThreshold
as a static data member.
Paul
--- ksynaptics-0.2.0.orig/ksynaptics/src/touchpad.cpp
+++ ksynaptics-0.2.0/ksynaptics/src/touchpad.cpp
@@ -31,6 +31,7 @@
TouchPad
On Monday 22 August 2005 19:58, Daniel Jacobowitz wrote:
On Sun, Aug 21, 2005 at 01:23:56AM +0100, Paul Brook wrote:
Package: amd64-libs
Version: 1.2
Severity: normal
Neither amd64-libs nor the biarch gcc-4.0 packages contain a 64-bit
libgcc_s.so.1
Hi Paul,
It should
Package: amd64-libs
Version: 1.2
Severity: normal
Neither amd64-libs nor the biarch gcc-4.0 packages contain a 64-bit
libgcc_s.so.1
-- System Information:
Debian Release: testing/unstable
APT prefers unstable
APT policy: (500, 'unstable')
Architecture: i386 (i686)
Shell: /bin/sh linked to
Package: bash
Version: 3.0-15
Severity: normal
Tags: patch
Command completion for LVM Physical Volume and Logical Volume names is
broken. For example lvdisplay /dev/tab gives:
awk: ^ syntax error
awk: {if ($2 ~ /^/dev//) print $2}
awk: ^ unterminated regexp
On Tuesday 16 August 2005 09:25, Michael Kerrisk wrote:
Hello Paul,
You may have found a fault in the man page, but I believe
the fault may be the converse of what you think.
...
The following note on the same page implies that this behavior is
intentional, and the documentation incorrect:
Package: apt-cacher
Version: 1.0.6
Severity: important
Running apt-cacher-cleanup.pl (as run by the cron script) gives the error
below:
/usr/share/apt-cacher/apt-cacher-cleanup.pl -q
Global symbol @stat requires explicit package name at
/usr/share/apt-cacher/apt-cacher-cleanup.pl line 183.
Package: manpages-dev
Version: 2.02-2
Severity: normal
Tags: patch
The clone(2) manpage incorrectly states that CLONE_PARENT_SETTID stores
the child TID. In fact it stores the parent TID (verified on a live
system).
The following note on the same page implies that this behavior is
intentional,
Package: autossh
Version: 1.3-3
Severity: normal
Tags: patch
Autossh doesn't handle -- properly when parsing commandling options. It
checks for and eats the first --, but it still parses the subsequent
commandline arguments.
Attached patch fixes this. Any arguemnts after -- are passed verbaitum
Package: freeglut
Followup-For: Bug #296242
This package now FTBFS on i386 with the same error. Alternative patch
below.
diff -u clean/freeglut_callbacks.c ./freeglut_callbacks.c
--- clean/freeglut_callbacks.c 2004-03-30 09:37:27.0 +0100
+++ ./freeglut_callbacks.c 2005-07-16
Rebuilding the kdepim source package, and installing the resulting lib*.deb
fixed the problem for me.
Paul
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Package: uptimed
Version: 1:0.3.3-2.2
Severity: normal
Tags: patch
The provided HTML headers for teh cgi output set the background color,
but not the text color. Thi smakes the output illegible when using an
inverse (while-on-black) color scheme.
Patch below fixes this.
Paul
---
71 matches
Mail list logo