Bug#792307: closed by Brian Potkin <claremont...@gmail.com> (Re: Bug#863974: hplip should not require systemd)
Correct, sorry, I've been running without any systemd components for such a long time that I forgot the details. Either way, systemd components are currently pulled in and activated (logind-systemd). I don't have a good example for Linux off the top of my head because I've removed systemd a long time ago but maybe an example from OS X (which seems to be the origin of quite a few concepts introduced with systemd) explains my general problem: the socket used for X11 is stored in a private tmp diretory which can't be accessed by other users, thus I can't su to another login and still use X11 programs. That's what breaks my workflow - I usually have two or three different logins active on the same desktop and private tmp directories break things for me sooner or later. Of course I can set up a shared directory accessible by all users but that's not the point. Plus the ever-growing list of tmpfs mount points is really getting to me. I know that ConsoleKit is no longer maintained but that's what I'm using right now because it's set up as a dependency. Maybe it would be possible to ditch all dependencies to "fast user switching" without systemd and go back to the old way of things where ownership of console devices is set to whoever logs into a local console when no other console is active. This way, folks who don't want Linux turned into something resembling Windows or OS X can work the way they're used to and all others can have systemd and all the things that come with it... Like I said, I'm more than happy to provide a patch for policykit that does all that dynamically, i.e. doesn't need hard dependencies to systemd but uses it when present, dynamically loading the systemd libs. But if there's no interest it would be a waste of time. I'd also be willing to step up as maintainer for ConsolKit if that helps. Or both. On 06/04/2017 11:05 AM, Simon McVittie wrote: On Sat, 03 Jun 2017 at 22:50:58 +0200, Christian Mueller wrote: (separate temp mount points for each user) which, apart from the incredible clutter in the list of mounted file systems, breaks my workflows (I need a single /tmp for all users). systemd-logind mounts a small tmpfs at /run/user/$uid for each concurrent user, as its way to implement XDG_RUNTIME_DIR without letting users cause denial of service by filling up /run. /tmp remains visible to all users. Just having a version of policykit-1 compiled without systemd dependencies would solve all our issues and it's a tiny little change in the rules file. The change is tiny, but the support burden is not. To be able to implement the policies that it provides, polkit needs a way to determine which users are logged-in, which of those logged-in users are local (getty, xdm etc. but not ssh), and which of those local users are on the active VT. Historically, that was implemented by ConsoleKit, which no longer has upstream maintainers[1], and does not appear to have Debian maintainers either. On Linux systems (with either systemd, sysvinit + systemd-shim or Upstart + systemd-shim) the replacement is systemd-logind. S [1] https://www.freedesktop.org/wiki/Software/ConsoleKit/
Bug#792307: closed by Brian Potkin <claremont...@gmail.com> (Re: Bug#863974: hplip should not require systemd)
Hi Brian, I just realized that my bug report, #792307, got merged with #863974 which may have the same underlying cause but a different interpretation of the results. Yes, just installing the systemd binary won't enable it as the active init system but another part of the dependency chain, libpam-systemd, already imports some of the systemd patterns (separate temp mount points for each user) which, apart from the incredible clutter in the list of mounted file systems, breaks my workflows (I need a single /tmp for all users). Debian always maintained that systemd would be optional and I would hope for a little more flexibility when it comes to [reasonable] requests to allow setting up desktop systems without having systemd bits and pieces getting in the way. Just having a version of policykit-1 compiled without systemd dependencies would solve all our issues and it's a tiny little change in the rules file. We would simply have two alternatives for policykit-1, one with and one without systemd. Or dynamic support at runtime. If there's anything I can do to help getting this implemented, especially the runtime detection, please let me know - I'm more than happy to put this together but only if I this is not a lost cause because nobody is interested in accepting this as a patch, anyway. Thanks, --Christian On 06/03/2017 09:57 PM, Debian Bug Tracking System wrote: This is an automatic notification regarding your Bug report which was filed against the policykit-1 package: #792307: policykit-1: There should be a variant of policykit-1 which doesn't depend on systemd It has been closed by Brian Potkin. Their explanation is attached below along with your original report. If this explanation is unsatisfactory and you have not received a better one in a separate message then please contact Brian Potkin by replying to this email.
Bug#834732: dkms compiler erro
For the sake of completeness, here's the dkms compiler error with kernel 3.16.7: ... CC [M] /var/lib/dkms/fglrx/15.12/build/firegl_public.o /var/lib/dkms/fglrx/15.12/build/firegl_public.c: In function ‘KCL_LockUserPages’: /var/lib/dkms/fglrx/15.12/build/firegl_public.c:3231:5: error: implicit declaration of function ‘get_user_pages_remote’ [-Werror=implicit-function-declaration] ret = get_user_pages_remote(current, current->mm, vaddr, page_cnt, 1, 0, (struct page **)page_list, NULL); ^ ...
Bug#712841: A bit late, but still an issue...
That would be great. You mean customise.conf, correct? Anything I can do to help? Thanks, --Christian P.S.: Thanks for all the effort you put into this - I'm using Debian on ARM devices for my mail/DNS/VPN/... server(s) since the old NSLU2 days and it all started on on cyrius.com :) On 05/05/2016 10:59 PM, Martin Michlmayr wrote: * Christian Mueller <ch...@mumac.de> [2016-04-29 22:20]: Long story cut short: if there is a way to identify TS-119 vs. TS-219 hardware, a fix would be very welcome. There's a configuration file on /dev/mtdblock5 which contains the device type. I'm starting to think the easiest solution would be to mount that device, check the device type and configure qcontrol accordingly.
Bug#712841: A bit late, but still an issue...
Hi all, I've just upgraded my trusty and apparently indestructable TS-119 to Jessie and stumbled over this bug. I've simply commented-out the line that increases the fan error count and the beeping has stopped but I see log entries which show qcontrol trying to set fan speeds due to temperature measurements. Long story cut short: if there is a way to identify TS-119 vs. TS-219 hardware, a fix would be very welcome. If not, I guess an option in /etc/default/qcontrol to simply disable any kind of thermal management involving fans would be nice, too (overtemp handling excluded, of course) Thanks, --Christian
Bug#794420: Test file
I tried to send the test file as base64-encoded email yesterday (71KB, around 16 A4 pages when printed) but it never showed up in the bug report; I guess it might have been too large. Do you have any other means to upload test files? -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#794420: ghostscript: ps2pdf13 produces incomplete files with PS files generated by Windows printer drivers
Package: ghostscript Version: 9.06~dfsg-2+deb8u1 Severity: important When setting up a Windows client to print through my main Debian box, I noticed that the test page came out incomplete - only the logo and one letter was printed. I tracked this down to CUPS using compatibility version 1.3 when converting the PS file to a PDF file (/usr/lib/cups/filters/pstopdf). This can be reproduced using ps2pdf13 and ps2pdf12 with the file I indent to attach to this bug report (hoping that I can attach files). I've changed the CUPS filter to use PDF compatibility version 1.2 and can now print from Windows - this might be an acceptable workaround until the underlying cause has been fixed but this is, of course, a different package. I'm still submitting this bug to ghostscript because this is where I believe it's ultimately located. The importance of this defect is, I believe, relatively high because it appears toprevent any Windows client from printing anything on a Debian 8.1 server, at least in situations as mine whether the printer doesn't speak Postscript and CUPS needs to handle the input file. Please stand by for the example file (the printer test page from Windows). -- System Information: Debian Release: 8.1 APT prefers stable APT policy: (700, 'stable'), (500, 'stable-updates'), (500, 'oldstable-updates'), (500, 'oldstable') Architecture: amd64 (x86_64) Foreign Architectures: i386 Kernel: Linux 3.16.7-ckt11-amd64-cm1.5 (SMP w/6 CPU cores) Locale: LANG=en_US.utf8, LC_CTYPE=en_US.utf8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Init: sysvinit (via /sbin/init) Versions of packages ghostscript depends on: ii debconf [debconf-2.0] 1.5.56 ii gsfonts1:8.11+urwcyr1.0.7~pre44-4.2 ii libc6 2.19-18 ii libgs9 9.06~dfsg-2+deb8u1 ghostscript recommends no packages. Versions of packages ghostscript suggests: ii ghostscript-x 9.06~dfsg-2+deb8u1 -- no debconf information -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#792307: policykit-1: There should be a variant of policykit-1 which doesn't depend on systemd
Package: policykit-1 Version: 0.105-8 Severity: normal Policykit has been around for a while and many packages required it long before systemd appeared on the scene (in my case, blueman, lightdm-gtk-greeter, colord, hplip to name a few). It's companion, ConsoleKit, now comes in two flavors, one being systemd-logind, the other the original ConsoleKit, and the dependencies across the Debian packages have been set up to depend on either one of them. What I'm proposing is to use a similar solution for policyit. For example, there could be two packages: systemd-policykit-1 and policykit-1, the former one compiled with --enable-systemd, the latter compiled with --disable-systemd. Or, since we're talking about a release already out, the packages could be named policykit-1 (with systemd) and policykit-1-nosystemd (without systemd). Or something around those lines - you get the picture :) -- System Information: Debian Release: 8.1 APT prefers stable APT policy: (700, 'stable'), (500, 'stable-updates'), (500, 'oldstable-updates'), (500, 'oldstable') Architecture: amd64 (x86_64) Foreign Architectures: i386 Kernel: Linux 3.16.7-ckt11-amd64-cm1.4 (SMP w/6 CPU cores) Locale: LANG=en_US.utf8, LC_CTYPE=en_US.utf8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Init: sysvinit (via /sbin/init) Versions of packages policykit-1 depends on: ii dbus 1.8.18-0+deb8u1 ii libc6 2.19-18 ii libglib2.0-0 2.42.1-1 ii libpam0g 1.1.8-3.1 ii libpolkit-agent-1-00.105-8 ii libpolkit-backend-1-0 0.105-8 ii libpolkit-gobject-1-0 0.105-8 policykit-1 recommends no packages. policykit-1 suggests no packages. -- no debconf information -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#640729: netbase: /etc/sysctl.d/bindv6only.conf should not be the default
Package: netbase Version: 4.45 Severity: important Tags: ipv6 I've just discovered that netbase now sets the system parameter bindv6only. This seems completely wrong to me, especially in the light that acceptance for IPv6 is still very low simply because there's next to no interoperability with IPv4. Luckily, at least servers can listen for IPv4 and IPv6 connections with a single socket, using :::x.x.x.x addresses for IPv4 clients. I've been busy changing my own code (tons of stuff) to do just that (i.e. listen to a single IPv6 socket which will also accept IPv4 connections) and now I find out that this is disabled per default in Debian...! The comment in /etc/sysctl.d/bindv6only says: # This is the default behaviour of almost all modern operating systems. This is the first time I hear about this. All boxes I'm dealing with (SunOS, HP-UX, AIX, various Linux distributions) have this questionable feature disabled and will accept IPv4 connections on an IPv6 socket. The only operating system I know of that doesn't (didn't?) have a dual mode IP stack is Windows ;) -- System Information: Debian Release: 6.0.2 APT prefers stable APT policy: (500, 'stable') Architecture: i386 (i686) Kernel: Linux 2.6.32-cm1.7-686 (SMP w/4 CPU cores) Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Versions of packages netbase depends on: ii initscripts 2.88dsf-13.1 scripts for initializing and shutt ii lsb-base3.2-23.2squeeze1 Linux Standard Base 3.2 init scrip Versions of packages netbase recommends: ii ifupdown 0.6.10 high level tools to configure netw netbase suggests no packages. -- no debconf information -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#639779: netbase: /etc/sysctl.d/bindv6only.conf should not be the default
Package: netbase Version: 4.45 Severity: important Tags: ipv6 I've just discovered that netbase now sets the system parameter bindv6only. This seems completely wrong to me, especially in the light that acceptance for IPv6 is still very low simply because there's next to no interoperability with IPv4. Luckily, at least servers can listen for IPv4 and IPv6 connections with a single socket, using :::x.x.x.x addresses for IPv4 clients. I've been busy changing my own code (tons of stuff) to do just that (i.e. listen to a single IPv6 socket which will also accept IPv4 connections) and now I find out that this is disabled per default in Debian...! The comment in /etc/sysctl.d/bindv6only says: # This is the default behaviour of almost all modern operating systems. This is the first time I hear about this. All boxes I'm dealing with (SunOS, HP-UX, AIX, various Linux distributions) have this questionable feature disabled and will accept IPv4 connections on an IPv6 socket. The only operating system I know of that doesn't (didn't?) have a dual mode IP stack is Windows ;) -- System Information: Debian Release: 6.0.2 APT prefers stable APT policy: (500, 'stable') Architecture: i386 (i686) Kernel: Linux 2.6.32-cm1.7-686 (SMP w/4 CPU cores) Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Versions of packages netbase depends on: ii initscripts 2.88dsf-13.1 scripts for initializing and shutt ii lsb-base3.2-23.2squeeze1 Linux Standard Base 3.2 init scrip Versions of packages netbase recommends: ii ifupdown 0.6.10 high level tools to configure netw netbase suggests no packages. -- no debconf information -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#585786: New dependencies pull in mdadm and mdsetup even when not needed
Since xfce depends on eject and eject depends on libdevmapper, the fix to 585786 causes mdadm and mdsetup to be installed when I try to update my Squeeze box. However, I don't use RAIDs or LVM and would rather not have the corresponding tools installed. This may not be very important but it's a nuisance... Wouldn't it be better to have mdadm and mdsetup depend on a specific version of libdevmapper? Or is this an issue with eject depending on libdevmapper? Thanks, --Christian -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#570283: Some more information...
Since the gcc version used to compile dash might have an impact, here's the version I'm currently using. If you need more information (like a complete list of all packages), please let me know. Regarding the kernel: It's a stock Debian kernel source with one file patched to initialize the eSATA controller in the SheevaPlug (arch/arm/mach-kirkwood/sheevaplug-setup.c) $ gcc -v Using built-in specs. Target: arm-linux-gnueabi Configured with: ../src/configure -v --with-pkgversion='Debian 4.4.2-9' --with-bugurl=file:///usr/share/doc/gcc-4.4/README.Bugs --enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --enable-shared --enable-multiarch --enable-linker-build-id --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.4 --program-suffix=-4.4 --enable-nls --enable-clocale=gnu --enable-libstdcxx-debug --enable-objc-gc --disable-sjlj-exceptions --enable-checking=release --build=arm-linux-gnueabi --host=arm-linux-gnueabi --target=arm-linux-gnueabi Thread model: posix gcc version 4.4.3 20100108 (prerelease) (Debian 4.4.2-9) Another idea: If you could provide debug symbols for the dash build coming with Debian Sequeeze I could debug the original binary. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#570283: runaway memory leak when processing DBOX2 CDK configure scripts
Package: dash Version: 0.5.5.1-3 Severity: important Hi all, this is a bit of a weird problem. I tried to compile a DBOX2 (tuxbox) image on a SheevaPlug and found dash would eat up all memory while running certain configure scripts. In order to find out what's going on, I compiled dash with - apt-get source dash - cd dash-0.5.5.1 - dkpg-buildpackage -rfakeroot -nc and link the resulting binary to /bin/dash for debugging purposes. However, the problem did not occur with the binary compiled this way, with or without debugging information (DEB_BUILD_OPTIONS set ot unset) In order to reproduce the problem, a rather lengthy procedure is required (on a SheevaPlug with Debian Sequeeze and development tools): # useradd dbox # su - dbox $ mkdir prj; cd prj $ mkdir tuxbox; cd tuxbox $ for i in apps boot driver hostapps cdk; do git clone git://gitorious.org/~seife/tuxbox-cvs/$i.git done $ cd cdk $ ./autogen.sh $ ./configure --with-cvsdir=/home/dbox/tuxbox \ --prefix=/home/dbox/tuxbox/dbox2 \ --with-boxtype=dbox2 \ --with-checkImage=rename \ --enable-fx2-c64emu \ --enable-fx2-lemm \ --enable-fx2-mines \ --enable-fx2-sudoku \ --enable-fx2-yahtzee \ --enable-fx2-solitair \ --enable-fx2-satfind \ --enable-dvbsnoop \ --with-procps $ make yadd-neutrino This may work fine (I remember seeing an OOM in dmesg but it was somewhere in the middle of the build and hard to evaluate so I rebuilt with /bin/sh pointing to /bin/bash.) It will take around 1-2 hours to bootstrap the cross-build environment and compile the base installation. The next step, however, seems easier to reproduce: $ ulimit -v 786432 $ make extra ( rm -rf ncftp-3.2.2 || /bin/true ) bunzip2 -cd /home/dbox/prj/tuxbox/cdk/Archive/ncftp-3.2.2-src.tar.bz2 | TAPE=- tar -x / ((for f1 in config.guess config.sub; do (for f2 in `find ncftp-3.2.2 -name / $f1`; do (test -e $f2 rm -f $f2 ln -s / /home/dbox/prj/tuxbox/cdk/Patches/$f1 $f2 echo updated $f2) done) / / done) || /bin/true) updated ncftp-3.2.2/sh/config.guess updated ncftp-3.2.2/sh/config.sub cd ncftp-3.2.2 \ AR=powerpc-tuxbox-linux-gnu-ar AS=powerpc-tuxbox-linux-gnu-as CC=powerpc-tuxbox-linux-gnu-gcc CXX=powerpc-tuxbox-linux-gnu-g++ NM=powerpc-tuxbox-linux-gnu-nm RANLIB=powerpc-tuxbox-linux-gnu-ranlib STRIP=powerpc-tuxbox-linux-gnu-strip CFLAGS=-pipe -Os CXXFLAGS=-pipe -Os LDFLAGS=-Wl,-O1 PKG_CONFIG_PATH=/home/dbox/prj/tuxbox/dbox2/cdkroot/lib/pkgconfig \ ./configure \ --build=armv5tel-unknown-linux-gnueabi \ --host=powerpc-tuxbox-linux-gnu \ --prefix= \ make clean all LD=powerpc-tuxbox-linux-gnu-ld \ make install DESTDIR=/home/dbox/prj/tuxbox/dbox2/cdkroot creating cache ./config.cache checking if you set and exported the environment variable CC... powerpc-tuxbox-linux-gnu-gcc checking for environment variable CFLAGS... -pipe -Os checking for environment variable CPPFLAGS... no checking for environment variable LDFLAGS... -Wl,-O1 checking for environment variable LIBS... no checking version of C library... Segmentation fault ^C This is where dash used up all of the memory and ended up with a SEGV because I set ulimit -v 786432. As mentioned above, trying to compile dash on the SheevaPlug resulted in a binary that did not show this particular behavior. Might be an issue with the gcc version used to compile the packages, or a cross-compile issue... -- System Information: Debian Release: squeeze/sid APT prefers testing APT policy: (500, 'testing') Architecture: armel (armv5tel) Kernel: Linux 2.6.32-cm1.0-kirkwood Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Versions of packages dash depends on: ii debianutils 3.2.2 Miscellaneous utilities specific t ii dpkg 1.15.5.6 Debian package management system ii libc6 2.10.2-2 GNU C Library: Shared libraries dash recommends no packages. dash suggests no packages. -- debconf information: * dash/sh: true -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#347720: Correction for description of the effects of this bug
I mixed two things up when I was submitting this bug. The effect of this bug is *not* that the module load order changes but that too many modules get loaded. The problem I had with this, apart from having a bloated initrd image, was that the USB driver on the affected box was causing severe hardware problems, blocking interrupts for other PCI cards, and the image created by mkinitrd would always load the USB driver module at boot time regardless of what I put into /etc/mkinitrd/mkinitrd.conf With the suggested fix, only those modules which are really needed at boot time are loaded. Here's a comparison of the modules loaded with and without the fix: With the fix: - modprobe -k piix modprobe -k pdc202xx_new modprobe -k vesafb /dev/null 21 modprobe -k fbcon 2 /dev/null modprobe -k unix 2 /dev/null modprobe -k megaraid modprobe -k aic7xxx modprobe -k sd_mod Without the fix: modprobe -k piix modprobe -k pdc202xx_new modprobe -k vesafb /dev/null 21 modprobe -k fbcon 2 /dev/null modprobe -k unix 2 /dev/null modprobe -k megaraid modprobe -k aic7xxx modprobe -k vesafb modprobe -k font modprobe -k sd_mod modprobe -k ide-cd modprobe -k ide-disk modprobe -k ide-generic modprobe -k psmouse modprobe -k evdev modprobe -k mousedev modprobe -k tsdev modprobe -k raid0 modprobe -k e100 modprobe -k 3c59x modprobe -k pci_hotplug modprobe -k ehci-hcd# -- one of those caused modprobe -k ohci-hcd# -- hardware problems modprobe -k rtc modprobe -k pcspkr modprobe -k parport_pc modprobe -k floppy modprobe -k lp -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#278028: The problem is reproduceable
Hello, I was able to reproduce this problem using the non-server variant, version 1.8.7-3, on my Debian Sarge system. I get only the first appointment as popup and as email, and no notification of the second one at all. Best regards, Christian Müller.