Package: apt-cacher
Version: 1.1
Severity: normal
Tags: patch
Problem the first:
The apt-cacher2 script sets the user ID first, and the group ID second.
This is backwards---setgid() may not work if the script dropped the
necessary privileges in the setuid() call.
Problem the second:
The
Package: zsnes
Version: 1.420-2.1
Severity: wishlist
The current i386 package runs without modification on an amd64 system
with ia32-libs installed. Would like to request that zsnes be made
available through apt to the amd64 arch.
-- System Information:
Debian Release: 4.0
APT prefers
Package: gbdfed
Version: 1.1-2
Severity: grave
I am running a Debian/etch system on amd64.
The program will map a window, and can even load a BDF font. As soon as I
attempt to edit a character, however, or read the online help---crash! I am
seeing not only segfaults, but also glibc complaints
Package: openoffice.org
Version: 2.0.4~rc3-1
Severity: important
I start up the program by invoking openoffice, wait for the main window
to come up, and then hit Ctrl-Q to exit.
The window stays up, but is no longer painted. The soffice.bin process
remains present, taking up memory but no CPU;
Package: dirdiff
Version: 2.1-1
Severity: important
An example: Create a file named xyzzy[foo bar].txt, and place it in a
directory tree so that dirdiff displays the file path. It is presented as
/path/to/bar].txt
/path/to/xyzzy[foo
Another example: Try a file named Foo Bar
Package: iceweasel
Version: 2.0.0.3-1
I was testing Content-Encoding: gzip of CSS and Javascript files on my
site. Iceweasel does not handle these at all.
With gzip-encoded CSS, the page loads as if the stylesheet could not be
downloaded at all. The Error Console gives Selector expected and
On Mon, 2007 May 14 08:40:22 +0200, Mike Hommey wrote:
Could you install mozilla-livehttpheaders, open tools-live http
headers, and check out the headers the server sends and post them here ?
Oh, f***... the server is sending Content-Encoding: gzip, gzip, which
apparently means that mod_gzip
Package: svgalib-bin
Version: 1:1.4.3-21
Severity: normal
The textmode(1) script starts off like so:
begin
#!/bin/sh
dir=/etc/vga
[ -w $dir ] || dir=$HOME/.vga
end
It checks whether or not /etc/vga is writable, even though it will only
be reading saved VGA state files. If
The latest xscreensaver has been activating on me in the middle of various
full-screen SDL games, when I use keyboard-only controls. The exact
behavior varies:
armagetron: Full-screen window minimizes (showing the desktop), screen
immediately fades to black, hack starts running
Package: thttpd
Version: 2.23beta1-4
Severity: minor
The config.h file has the line
#define STATS_TIME 3600
which causes the following sort of blurb to be written to daemon.log once
every hour:
begin log excerpt
... thttpd[11862]: up 3600 seconds, stats for 3600 seconds:
...
I've sent Jamie a patch that fixes the /proc/interrupts detection. The next
version of XScreenSaver---which, according to him, will be 5.00---should do
away with this bug.
--Daniel
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL
On Thu, 2006 May 04 09:38:11 +0200, Ralf Hildebrandt wrote:
And how about sending it here, too?
Oh, of course. Attached.
This is against 5.00b4, but it applies cleanly to 4.24 as well. It's a
pretty trivial patch, as you can see
--Daniel
--- xscreensaver-5.00b4/configure.in
Package: nut
Version: 2.0.3-4
Severity: important
Tags: patch
When the init script's poweroff command is called from
/etc/init.d/halt (via ups-monitor poweroff): If the UPS does not cut
power immediately, control will return to the halt script, which will
subsequently power off the system.
This
I'm seeing this same bug on installation of some CJK fonts. (This is on
Ubuntu/dapper, but it may still provide a useful data point.)
# apt-get install ttf-kochi-gothic ttf-arphic-gkai00mp ttf-baekmuk
[snippage]
Unpacking ttf-kochi-gothic (from .../ttf-kochi-gothic_1.0.20030809-3_all.deb)
...
Same situation here as Tony's (Debian + Ubuntu), using path_map. My
solution was to run two separate instances of apt-cacher, but having one
handle both would be preferable.
Now I can change the code a bit to not remove the parent directories
when storing the files. However the next problem
On Tue, 2006 Apr 04 22:33:46 +0200, Eduard Bloch wrote:
Moving to directories carrying the download origin in the path, or
encoding the same information into the filename - that really does not
make a big difference. In contrary, And I can also imagine a worst case
where the encoded name can
On Wed, 2006 Apr 05 09:00:04 +0200, Eduard Bloch wrote:
So, finally, I decided to take the bitter pill and rewrite it in C++. I
think the implementation is done to ~60 percent now, including partial
component tests. From the current POV I think this will be a
double-threaded daemon, not
Package: apt-cacher
Version: 1.5.3
Severity: important
Ran into this when attempting to install Debian sarge off an apt-cacher
server. The installer invokes busybox-wget to grab the Release file,
yielding the error message wget: no response from server. The error is
reproducible with a
On Fri, 2006 Jul 28 14:46:47 -0300, Henrique de Moraes Holschuh wrote:
1. The UPS may take more than 15 minutes to shutdown the load. You cannot
assume things like this, and you will cause data loss if you get it wrong:
the power-off could come with the system fully online.
The time period
On Fri, 2006 Jul 28 16:12:35 -0300, Henrique de Moraes Holschuh wrote:
There is no tradeoff without the hack, and the hack is only needed in
hardware unsuitable for UPS management. Thus, it must be optional. It is
dangerous to data and the hardware, so it should not be the default.
Define
On Mon, 2006 Jul 31 13:47:21 -0300, Henrique de Moraes Holschuh wrote:
Define (un)suitable for UPS management. Does this definition include
most people's desktop systems?
Suitable for UPS management:
Load:
Powers up when AC returns
Can be informed
Package: xapple2
Version: 0.7.4-5
Severity: important
On a modern X11 setup with TrueColor/DirectColor displays, xapple2 refuses
to start, giving the following error:
$ xapple2
Sorry bud, xapple2 only supports 8bit PseudoColor displays.
Maybe you can fix this!
This
Package: openoffice.org
Version: 2.0.4.dfsg.2-7etch4
After the initial program window comes up: Between clicking the File menu
for the first time, and seeing the menu pop up, is a long delay---on the
order of 20 seconds---during which neither the keyboard nor mouse can
interact with OO.org nor
Package: gbdfed
Version: 1.2-1
This is on a Debian/testing system on amd64.
Activating any of the items under the top-level Help menu causes the
program to segfault instantly.
The problem is at least partly due to the use of an incorrect terminating
sentinel in calls to
Mark,
There is a bug report on this same issue on Launchpad; it has a bit of
background that may be informative:
https://bugs.launchpad.net/bugs/366293
Note that the Translation file that apt-get(8) attempts to download is
dependent on the current value of LANG. I have LANG=en_US.UTF-8, and
Package: libxslt1-dev
Version: 1.1.26-1
The -dev package lacks the libxslt.m4 Autoconf macro file. (Note that
the libxml2-dev package does include libxml2.m4.)
This bug is mirrored in the Ubuntu bug tracker:
https://bugs.launchpad.net/bugs/287342
--
To UNSUBSCRIBE, email to
I've investigated this bug further, and found the problem. Here is a
walk-through of it, in apt-cacher 1.6.8:
1. First, in the handle_connection() subroutine (line 243), you read in
the request. apt-get(8) typically uses HTTP Keep-Alive for efficiency
(i.e. download multiple files in one
I believe Bug#517761 is a duplicate of this one, if I've diagnosed the
problem correctly. Please have a look at
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=517761#69
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact
I've investigated this bug, in apt-cacher 1.6.8, and believe I have
found the problem.
First, here is a typical manifestation, from apt-get(8) downloading a
slew of packages for the xubuntu-desktop metapackage:
8
Get:600 http://midnight jaunty/main python-rdflib 2.4.0-5ubuntu1 [302kB]
Package: apt-cacher
Version: 1.6.8
Severity: wishlist
Tags: patch
Currently, a single instance of apt-cacher cannot serve both Debian and
Ubuntu systems, even though it could serve either one alone. The reason
for this is that there are numerous package files that have the exact
same filename in
8
Hit http://midnight jaunty-security Release.gpg
Ign http://midnight jaunty-security/main Translation-en_US
Err http://midnight jaunty-security/restricted Translation-en_US
Error reading from server - read (104 Connection reset by peer)
Ign http://midnight jaunty-security/universe
On Sun, 2009 Aug 16 21:01+0100, Mark Hindley wrote:
Thanks for this. Interesting. When I originally went through sorting
out the 2 server workaround for this, I considered (and rejected :))
this approach.
The problem with it is that apt-cacher was designed to ignore the
server path so that
On Sun, 2009 Aug 16 21:10+0100, Mark Hindley wrote:
OK. Thanks for this. I have been unable to reproduce.
Could you try with LANG=en_US.UTF-8? (And you have pipelining enabled,
and are using path_map, yes?)
I can reproduce this bug in a minimal en_US Lenny install, with a hand-
built
On Mon, 2009 Aug 17 09:11+0100, Mark Hindley wrote:
Lying awake considering this more overnight, I realised this would
also break checksumming which only uses the filename without a path.
Well, for one, you wouldn't lop off the directory parts of the filename
in apt-cacher-lib.pl:268 :-)
That
On Mon, 2009 Aug 17 19:44+0100, Mark Hindley wrote:
I meant, if you weren't using the path_map at all. Currently, if one
apt client requests
ftp.uk.debian.org/debian/pool/main/z/zip/zip_2.32-1_amd64.deb
and one
ftp.us.debian.org/debian/pool/main/z/zip/zip_2.32-1_amd64.deb
through the
On Mon, 2009 Aug 17 20:00+0100, Mark Hindley wrote:
Consider the cached file
/var/cache/apt-cacher/packages/zip_2.32-1_amd64.deb which is present
in an existing cache and also present in both Debian and Ubuntu
distros (different files, but same name). There is no 'long filename'
file, so
On Mon, 2009 Aug 17 19:24+0100, Mark Hindley wrote:
I still don't see it. But I am happy with your analysis. My default
LANG is en_GB, so I would have expect to have seen it with that as
that is also a missing Translation file
I'd feel better if you were able to reproduce it, but if you're
On Mon, 2009 Aug 17 20:58+0100, Mark Hindley wrote:
I am concerned that there might be some users on low end systems with
poor bandwidth and/or disk space who might consider the traditional
lack of redundant downloads as a real feature. So, since a multi-
distro cache can't use short-filename
On Mon, 2009 Aug 17 21:25+0100, Mark Hindley wrote:
I'll gladly put one together, but what should the behavior be? My
preliminary patch had it not closing the connection for 404 errors.
Should that be broadened to all errors (in which case, the whole if-
conditional can be removed), or
On Mon, 2009 Aug 17 21:35+0100, Mark Hindley wrote:
Actually, there is one more bit that concerns me. That is the checksum
code. Do you use checksumming?
I don't use checksumming, but aren't the checksum databases built off
the index files, not the packages?
At the moment there is a set of
On Mon, 2009 Aug 17 23:31+0100, Mark Hindley wrote:
It isn't that there would be multiple databases, just that the size
would/could increase dramatically.
Currently each .deb has it's checksums in the database. About 30MB for
the Debian distro. No hosts or paths are stored. Filenames are
Okay, here is a revised patch implementing the feature. This time,
apt-cacher-cleanup.pl and the checksum database are onboard. I also
tweaked the long-filename convention to use :: right before the actual
package name, to make the names easier to parse in Perl. (Colons can
also appear in the
On Wed, 2009 Aug 19 22:50+0100, Mark Hindley wrote:
5. If no, complain that the package can't be imported.
Does this sound reasonable? #5 would be a new behavior for the
import script, but these would be files that apt-cacher-cleanup.pl
would blow away anyway.
I think i would leave
Dominique,
The known workarounds at this point are as follows:
1. (Client-side) Set LANG=C when running apt-get update.
2. (Server-side) Apply Mark's patch in message #125, and if you are
using path_map, edit it so that each map component has only one
server behind it.
--
To
Okay, I think I've found out what was going on with server-hopping!
Look at the libcurl() routine, starting on line 1297 of version 1.6.8.
In there you have the bit that interacts with the curl download thread,
notably the while($libcurl) loop starting at line 1331.
First, you read in the HTTP
On Fri, 2009 Aug 21 10:55+0100, Mark Hindley wrote:
I will roll all these (apart from the multiple distro) and get 1.6.9
uploaded.
Yay! That'll be great, to get all these changes checkpointed.
It's possible that the remaining Connection reset by peer error
messages I'm seeing are not actual
Package: apt-cacher
Version: 1.6.8
This bug report is a spinoff of Bug#517761.
In the course of investigating that bug, I decided to set up a
simulated Debian repository so that I could stress-test apt-cacher's
handling of apt-get update without real Debian repository mirrors
seeing the load.
On Fri, 2009 Aug 21 10:35-0400, Daniel Richard G. wrote:
I've set up a Web server to simulate a repository, and now I'm seeing
yet another bug: both apt-cacher and apt-get come to a point where
each is expecting network traffic from the other, but nothing ever
comes---a sort of network
On Mon, 2009 Aug 24 21:42+0100, Mark Hindley wrote:
I have just been going through the RFCs. Even if thttpd is only doing
HTTP/1.0, it says
Applications should use this field to indicate the size of the Entity-
Body to be transferred, regardless of the media type of the entity. A
valid
Package: lynx-cur
Version: 2.8.7dev9-2.1
I preseed the debconf database with a value to use for lynx-cur/defaulturl.
When I install lynx-cur, and the debconf question comes up (Please enter
the default URL to use if none is given...), the default value given is
not the preseeded URL, but the
The lvm2 package in Ubuntu Intrepid has an identical config file (only the
accept every block device filter is uncommented), yet vgscan(8) does not
exhibit this behavior.
# vgscan
Reading all physical volumes. This may take a while...
#
--
To UNSUBSCRIBE, email to
Package: libhesiod0
Version: 3.0.2-20
Preseeding the debconf database with a value for hesiod/rhs does not
work. When the package is installed, it unconditionally uses the
system's domain (as given by the search directive in /etc/resolv.conf)
as the default value.
Using the net domain as a
Package: openbox
Version: 3.4.11.1-1
A default installation of openbox includes a root menu with an ObConf
item, which fails to do anything because obconf is not installed,
because obconf is suggested and not recommended by openbox, so obconf
doesn't normally get pulled in when the user installs
On Fri, 2010 Sep 24 22:01+0200, Nico Golde wrote:
Sorry but I disagree, most users are probably not using this tool.
I would be quite surprised if most users prefer to edit the
configuration by hand, instead of using a reasonably polished GUI tool.
That said, if obconf is judged not to be
On Sat, 2010 Sep 25 00:41+0200, Nico Golde wrote:
I can only guess in this case. I may be totally wrong. What makes me
think this way is that it's not that actively maintained upstream and
can't configure quite a lot of features openbox provides.
That is a fair point. I was trying out openbox
The workaround for this bug is to add the following lines to
/etc/modprobe.d/blacklist:
blacklist agpgart
blacklist intel-agp
With that in place, the system boots and operates perfectly, without
needing any of the other kernel parameters.
Note: I did test with agp=off, and
Package: installation-reports
Severity: important
Installing debian-testing-i386-businesscard.iso, dated 17-Dec-2008, on a
Dell Dimension 2350 with an add-on PCI video card.
There is an existing kernel bug (see bug 507955) that causes the newly
installed Debian system to oops and become
Ping.
DKMS is used nowadays in the Ubuntu packaging of VirtualBox to keep the
associated kernel module in sync with the kernel. (Have a look-see at
virtualbox-ose-source.) It works very well there, and I think it is worth
investigating as a potential solution to this bug.
The only possible
On Wed, 2009 Feb 18 21:50:19 -0800, Russ Allbery wrote:
I've already finished DKMS packages for OpenAFS and will upload them as
soon as DKMS is available in Debian.
Russ, very kind thanks for the work you've done for OpenAFS :-)
I hope these new DKMS-enabled packages can find their way into
Package: mdadm
Version: 2.6.7.2-1
I set a value of none for mdadm/initrdstart using debconf-set-selections,
prior to installing mdadm. (I also mark the question as seen false, so
that the value can be changed if needed at package-install time; the intent
is simply to change the default.)
When
On Tue, 2009 Feb 24 20:30:30 +0100, martin f krafft wrote:
I forget debconf details, but does this work for other settings for
you? I.e. are you sure that seen false doesn't cause it to ignore
the preseeded value and use the default anyway?
Positive. For most packages, if the question comes
Package: fontconfig-config
Version: 2.8.0-2.1
Severity: minor
The fontconfig/enable_bitmaps debconf question (Enable bitmapped fonts
by default?) is asked twice when the package is apt-get-installed:
first at the Preconfiguring packages stage, and again at the Setting
up stage.
--
To
Package: gdb-minimal
Version: 7.0.1-2+b1
Severity: wishlist
Several debugger frontend packages specify gdb either in their
Depends: (e.g. xxgdb) or Recommends: (e.g. ddd). No such package is
aware of gdb-minimal. If I have gdb-minimal installed, and I install
one of these debugger frontends, APT
A new set of squeeze debian-installer images has appeared in the
repositories (dated July 6 for i386, July 10 for amd64), and with these,
the entire installation process functions flawlessly.
Bug #587570, I believe, can be closed. (I can't speak for bug #588183.)
--
To UNSUBSCRIBE, email to
Package: debian-installer
Version: 20100211+b1
(Note: I'm not sure that d-i is the correct package for this bug; please
reassign if needed.)
I want to install a minimal Debian/squeeze system. When the installer
presents the tasksel screen, I unselect everything, even Standard
system utilities.
On Thu, 2010 Jul 15 22:41+0200, Petter Reinholdtsen wrote:
Why do you find exim4 inappropriate for a minimal install?
* Users may not want a full MTA installed in the first place. (Same
argument as to why Apache/ProFTPD/NTPD/Bind/etc. are not appropriate
for a minimal install.)
* As a
On Sat, 2010 Jul 17 09:56+0200, Petter Reinholdtsen wrote:
The obvious fix is to change popularity-contest to suggest instead of
recommend, to avoid pulling in a MTA when popularity-contest is
installed. Cc to the maintainer list to see if that is an interesting
alternative.
Hunh. I wasn't
On Sat, 2010 Jul 17 13:35+0200, Frans Pop wrote:
So: as the reported issue is already fixed in current daily built D-I
images and as the Recommends in popcon is functionally correct, there
is absolutely zero need to make any changes in popcon.
Okay, now I'm a little confused.
So I see in the
On Sat, 2010 Jul 17 21:40+0200, Frans Pop wrote:
I'm not sure what image was used by the bug reporter, but I assume a
current D-I alpha 1 image. The alpha1 images date from mid Februari.
These received a refresh a few days ago:
On Sun, 2010 Jul 18 04:46+0200, Frans Pop wrote:
No, that's no update. Only a date change for some reason. The images
are still identical to the alpha1 release from February. One way to
see that is to check the kernel version: it uses 2.6.30 instead of
2.6.32. Another is to check the
On Sun, 2010 Jul 18 06:09+0200, Frans Pop wrote:
No. Those are *not* daily built images. Follow the relevant links from
the page I referred to! For netboot you need other images.
Oh, right---I see the different set of links, that go to various places
(e.g. people.debian.org) not in the
Package: dictionaries-common
Version: 1.5.5
Severity: minor
I saw this in my unattended-upgrades package installation log:
begin excerpt
Setting up libkrb5-dev (1.8.1+dfsg-2ubuntu0.2) ...
Setting up hunspell-en-ca (1:3.2.0-3ubuntu3.1) ...
Use of uninitialized value $ENV{HOME} in
On Fri, 2010 Jul 23 14:26+0200, Agustin Martin wrote:
This is a bit strange, Is HOME undefined at your site? Or may be it is
nil'ed during unattended upgrades. If the last, I will make it
optional, it is used only for $userdefault which is only useful for
interactive use.
I'm pretty sure
Package: glibc-doc
Version: 2.11.2-2
The (e)glibc manual has the following passage:
@comment math.h
@comment BSD
@deftypefun int finite (double @var{x})
@comment math.h
@comment BSD
@deftypefunx int finitef (float @var{x})
@comment math.h
@comment BSD
@deftypefunx int finitel (long double
Package: fontconfig-config
Version: 2.8.0-2.1
I am using debconf-set-selections(1) to preseed configuration settings
on a Debian sid system. One of these addresses Fontconfig's subpixel-
rendering feature; I want to completely disable it:
fontconfig-config fontconfig/subpixel_rendering
I reported this bug a while ago in Ubuntu:
https://bugs.launchpad.net/debian/+source/fontconfig/+bug/262089
However, it ceased to be an issue for them because they got rid of the
debconf questions altogether. (Is there any desire to harmonize the
fontconfig packages between Debian and
Package: texlive-extra-utils
Version: 2009-7
Severity: wishlist
xindy, unlike most TeX/LaTeX-related programs, requires Common Lisp.
Debian's packaging of clisp has lots of debconf questions and is fairly
cumbersome to be just another anonymous dependency. I think it would be
more appropriate to
Package: sysvinit
Version: 2.88dsf-9
(Note: This bug may also be applicable to the initscripts package,
which owns the /etc/default/halt config file.)
Normally, when /etc/default/halt specifies HALT=poweroff, invoking
halt or poweroff results in the system powering off. I want the more
Package: debian-installer
Version: 20100211
Severity: grave
The installers are currently broken due to bug 583551, which prevents
mkfs.ext3 et al. from working. New images are needed, per tytso in
this message:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=583551#117
Please ensure that
Probably would be useful to open a bug against the debian-installer to
get them to update their images? The relevant versions of e2fsprogs
are now (finally!) migrated into testing...
Bug #587570.
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of
Package: virtualbox-ose-qt
Version: 3.2.6-dfsg-2
I am using VirtualBox on a corporate network with a mandatory (but not
transparent) HTTP proxy. My environment has the http_proxy environment
variable set appropriately, and most applications work fine with that.
But when VirtualBox goes to
Package: chrony
Version: 1.23-7
Severity: minor
When I install chrony, I get the following message in my inbox:
The password for chronyc is in /etc/chrony/chrony.keys.
Chrony has been configured to assume that your real-time clock is on UTC
time. If this is not correct edit
Package: libhesiod0
Version: 3.0.2-20
Severity: important
When libhesiod0 is installed with debconf at low priority, it queries
the user for hesiod/rhs, with a default value that uses the system's
currently configured DNS domain. If I enter a different value, this
value is not used---debconf and
Package: xdm
Version: 1:1.1.10-3
I installed a squeeze system for testing purposes, with X and a desktop
environment, and xdm as the display manager. The hostname is
test-linux, but for some odd reason, the banner displayed by xdm is
Welcome to test-1.
--
To UNSUBSCRIBE, email to
Package: debian-installer
Version: 20100805-15:58
Severity: important
I am attempting to install lenny with GRUB2 using a daily build of d-i.
After being prompted whether or not to install to the MBR, I get a red-
screen error as follows:
[!!] Configuring grub-pc
Unable to install GRUB
On Fri, 2009 Sep 11 10:22+0100, Mark Hindley wrote:
I thought I had worked out a locking error that was causing this, but
obviously not. Could you remove these last patches and turn on
debugging and send the error.log so I can try to see why $cached_file
is not getting initialised. As you
On Sat, 2009 Sep 19 21:19+0100, Mark Hindley wrote:
Thanks, very helpful. I think I have found this.
Can you verify it is fixed with this patch
I've been running the patched apt-cacher since Saturday evening in the
same 3-minute testing cycle, and with access.log now over 100K lines,
On Tue, 2009 Aug 25 09:22+0100, Mark Hindley wrote:
Could you try this patch
Okay, so all files with errors (including 404s) are sent as body-less,
Content-Length: 0 responses.
This does get rid of the timeouts for the 404 pages, which
addresses thttpd's limitations; apt-get update works
On Tue, 2009 Aug 25 21:13+0100, Mark Hindley wrote:
But what about servers that don't return Content-Length even for
successful requests?
This is difficult. I don't think it can have happened really, as I
notice return_file() relies on Content-Length being set to initialise
$explen. It
On Wed, 2009 Aug 26 11:28+0100, Mark Hindley wrote:
I think this is a buffering issue.
I can fix it here, and get the right size with the patch below. The
other alternative is to use autoflush on $$pkfdref. I need to do some
testing to see if that is a better option.
The patch does the
Package: apt-cacher
Version: 1.6.8
I am testing apt-cacher with checksumming enabled.
I start the server with an empty cache directory tree and no database
file (such that the database has to be created from scratch), and then
run apt-get update on two client machines (in this case, an Ubuntu
Package: apt-cacher
Version: 1.6.8
I am testing apt-cacher with checksumming enabled.
I start the server with an empty cache directory tree and no database
file (such that the database has to be created from scratch), and then
run apt-get update on two client machines (in this case, an Ubuntu
On Wed, 2009 Aug 26 22:26+0100, Mark Hindley wrote:
My understanding is that if perl itself blows up like this, it is a
perl, or more-likely in this case libberkeleydb bug.
I suggest resumbmitting against that package. Or I can reassign
Better to reassign, I think. At this stage, the bug
On Thu, 2009 Aug 27 11:19+0200, Marco d'Itri wrote:
This means that a malloc data structure is damaged. It may be caused
by BerkeleyDB.xs or not, probably not. To get *any* help you need to
provide useful information, start by running the program in valgrind
to have a chance of finding out
On Fri, 2009 Aug 28 01:02+0100, Mark Hindley wrote:
Actually, I think it should be fixed by not releasing the lock until
the end of the fetcher process.
I'm still seeing occasional bzip2 errors with this patch. (I have a good
test rig for this now: four clients apt-get update-ing from a server
On Tue, 2009 Sep 1 08:39+0100, Mark Hindley wrote:
I was holding off, because I wanted this fixed in it!
Ah, okay. From the look of the control messages, I thought you'd
uploaded it already.
I have rewritten the decompression code to use native perl modules
rather than external command
After a few hundred cycles of testing, this patch is looking solid.
Running with checksumming enabled, and debugging disabled, error.log and
db.log remain at zero size.
Oh, but there was one minor issue. I was previously seeing this warning
come up repeatedly in error.log:
Tue Sep 1 18:10:46
On Sun, 2009 Sep 6 10:58+0100, Mark Hindley wrote:
Are the types of file that $cache_status is uninitialised for all
package files, all index files or a mixture?
I was only testing downloads of index files, so no word on whether the
warning would have come up for package files as well.
--
On Sun, 2009 Sep 6 11:41+0100, Mark Hindley wrote:
OK. Is it 1 type of index file?
From the looks of it, it's all types:
$ grep UNKNOWN access.log | sed 's/.*_//' | sort -u
Packages.bz2
Release
Release.gpg
Sources.bz2
(UNKNOWN was what I was initializing $cache_status with.)
--
To
On Thu, 2009 Sep 10 11:43+0100, Mark Hindley wrote:
Could you try this patch and see if it is fixed for you.
Sorry, that was incomplete. You will also need this:
I'm still seeing some uninitialized value warnings, as well as a new
flock() warning cropping up in spades. Here's what error.log
1 - 100 of 279 matches
Mail list logo
