[Fink-devel] Apt cache limit
Recently, apt-get started crashing on me with the following error message:
E: Dynamic MMap ran out of room
E: Error occured while processing system-pkgconfig-xcb-property (NewPackage)
E: Problem with MergeList /tmp/finkaptstatus
The incriminated package does not matter, during my tests it was not
always the same, but the error came up persistently, whenever anything
apt-get related was run, apt-get or apt-cache directly, or fink
scanpackages or selfupdate etc.
After some googling, it is clear that this is a real problem related to
the increasing size of the package data bases and perhaps to the number
of repositories mentioned in sources.list. On google I also found a
workaround which might be interesting for others, as I am sure others
will be hit by this, too:
Create a file /sw/etc/apt/apt.conf.d/30cache containing one line
APT::Cache-Limit 125829120;
This sets the limit to 120MB instead of the ridiculously low default
6MB. Even 12MB would suffice.
A permanent fix would involve a patch to the file apt-pkg/pkgcachegen.cc
in the apt sources as follows:
--- apt-0.5.4/apt-pkg/pkgcachegen.cc~ 2009-12-02 14:16:11.0 +0100
+++ apt-0.5.4/apt-pkg/pkgcachegen.cc2009-12-02 14:18:17.0 +0100
@@ -599,3 +599,3 @@
{
- unsigned long MapSize = _config-FindI(APT::Cache-Limit,6*1024*1024);
+ unsigned long MapSize = _config-FindI(APT::Cache-Limit,60*1024*1024);
There is another line in that file where MapSize is even lower at 4MB,
but that does not seem to matter as much.
--
Martin
--
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
___
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
http://news.gmane.org/gmane.os.apple.fink.devel
Subscription management:
https://lists.sourceforge.net/lists/listinfo/fink-devel
Re: [Fink-devel] Apt cache limit
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Martin Costabel wrote:
Recently, apt-get started crashing on me with the following error message:
E: Dynamic MMap ran out of room
E: Error occured while processing system-pkgconfig-xcb-property (NewPackage)
E: Problem with MergeList /tmp/finkaptstatus
The incriminated package does not matter, during my tests it was not
always the same, but the error came up persistently, whenever anything
apt-get related was run, apt-get or apt-cache directly, or fink
scanpackages or selfupdate etc.
After some googling, it is clear that this is a real problem related to
the increasing size of the package data bases and perhaps to the number
of repositories mentioned in sources.list. On google I also found a
workaround which might be interesting for others, as I am sure others
will be hit by this, too:
Create a file /sw/etc/apt/apt.conf.d/30cache containing one line
APT::Cache-Limit 125829120;
This sets the limit to 120MB instead of the ridiculously low default
6MB. Even 12MB would suffice.
A permanent fix would involve a patch to the file apt-pkg/pkgcachegen.cc
in the apt sources as follows:
--- apt-0.5.4/apt-pkg/pkgcachegen.cc~ 2009-12-02 14:16:11.0 +0100
+++ apt-0.5.4/apt-pkg/pkgcachegen.cc 2009-12-02 14:18:17.0 +0100
@@ -599,3 +599,3 @@
{
- unsigned long MapSize = _config-FindI(APT::Cache-Limit,6*1024*1024);
+ unsigned long MapSize = _config-FindI(APT::Cache-Limit,60*1024*1024);
There is another line in that file where MapSize is even lower at 4MB,
but that does not seem to matter as much.
Ah. I got the same thing, myself. Fortunately in my case the offending
package was a buildlock, so I just deleted it.
- --
Alexander Hansen
Fink User Liaison
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.4.10 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
iEYEARECAAYFAksWiMIACgkQB8UpO3rKjQ8B9gCfdAjoqQpias9lOR0iHz53r1tc
VEoAn2rDPinPP5X2p/qBUB3s6cnGUJwR
=2OAA
-END PGP SIGNATURE-
--
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
___
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
http://news.gmane.org/gmane.os.apple.fink.devel
Subscription management:
https://lists.sourceforge.net/lists/listinfo/fink-devel
Re: [Fink-devel] Apt cache limit
Alexander Hansen wrote: [] Ah. I got the same thing, myself. Fortunately in my case the offending package was a buildlock, so I just deleted it. I am curious how long it will take you to see it again. I don't think this can be solved by removing one package. It is like when your harddisk is full: You can delete a file to buy a little time, but the problem will soon come back. -- Martin -- Join us December 9, 2009 for the Red Hat Virtual Experience, a free event focused on virtualization and cloud computing. Attend in-depth sessions from your desk. Your couch. Anywhere. http://p.sf.net/sfu/redhat-sfdev2dev ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-devel
Re: [Fink-devel] Apt cache limit
Martin Costabel wrote: Alexander Hansen wrote: [] Ah. I got the same thing, myself. Fortunately in my case the offending package was a buildlock, so I just deleted it. I am curious how long it will take you to see it again. I don't think this can be solved by removing one package. It is like when your harddisk is full: You can delete a file to buy a little time, but the problem will soon come back. I think the files to watch are /sw/var/cache/apt/(src,}pkgcache.bin. If they get bigger than 6MB (= 6291456 Bytes), you are in trouble. If the size is just below, you will be in trouble soon. Adding the unofficial tokyo bindist (for 10.5) to /sw/etc/apt/sources.list adds a little over 4MB to the cache files. -- Martin -- Join us December 9, 2009 for the Red Hat Virtual Experience, a free event focused on virtualization and cloud computing. Attend in-depth sessions from your desk. Your couch. Anywhere. http://p.sf.net/sfu/redhat-sfdev2dev ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-devel
Re: [Fink-devel] Apt cache limit
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Martin Costabel wrote: Martin Costabel wrote: Alexander Hansen wrote: [] Ah. I got the same thing, myself. Fortunately in my case the offending package was a buildlock, so I just deleted it. I am curious how long it will take you to see it again. I don't think this can be solved by removing one package. It is like when your harddisk is full: You can delete a file to buy a little time, but the problem will soon come back. I think the files to watch are /sw/var/cache/apt/(src,}pkgcache.bin. If they get bigger than 6MB (= 6291456 Bytes), you are in trouble. If the size is just below, you will be in trouble soon. Adding the unofficial tokyo bindist (for 10.5) to /sw/etc/apt/sources.list adds a little over 4MB to the cache files. Ah, that would make sense; they don't purge older versions. - -- Alexander Hansen Fink User Liaison -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.10 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAksWp8QACgkQB8UpO3rKjQ9kbQCeMJuUh1AUw/Ezr53F270vczgD J/UAn2m4gsgN4nv8xZR2mshbR2u5hyZM =y60L -END PGP SIGNATURE- -- Join us December 9, 2009 for the Red Hat Virtual Experience, a free event focused on virtualization and cloud computing. Attend in-depth sessions from your desk. Your couch. Anywhere. http://p.sf.net/sfu/redhat-sfdev2dev ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-devel
Re: [Fink-devel] New libgettext8 package
Am 01.12.2009 um 15:58 schrieb Jean-François Mertens: On 01 Dec 2009, at 15:18, Charles Lepple wrote: On Tue, Dec 1, 2009 at 8:44 AM, Jean-François Mertens j...@core.ucl.ac.be wrote: I notice there is buildconflicts for ccache-default : this breaks the operation of ccache on anything _ fink or not! _ going on in parallel.. An export CCACHE_DISABLE=1 (instead of the current env CCACHE_DISABLE=1, so as to apply also to the make command, not only the configure command) should suffice in principle ... Jean-François, After seeing your email, I went back to re-install ccache-default, and it looks like fink did it automatically. I remember seeing the message saying fink was going to temporarily remove it, and I guess it put it back when it was done. Did ccache-default remain uninstalled on your system? No ; in general, a BuildConlicts gets reinstalled by fink at the end of the build. That's why I mentioned above things going on in parallel. Exactly. I am unhappy whenever a package forces me to uninstall ccache- default, even if it is only temporary and and it gets reinstalled later on automatically: It affects all my concurrent build activities, which is negative. Also, I wonder what happens if I try to fink build two packages *in parallel* (i.e. in different terminals) that both conflict with ccache-default , esp. if one finishes earlier than the other ? Ideally, no package would ever build conflitc with ccache-default -- for most of them, it should be possible to set CCACHE_DISABLE (as Jean- François explained). This solution is far superior to a build conflict. It's faster (no need to remove and re-install another package), requires less user interaction, does not affect other ongoing compilations (be it by Fink or something else), i.e., is more self-contained. Bye, Max -- Join us December 9, 2009 for the Red Hat Virtual Experience, a free event focused on virtualization and cloud computing. Attend in-depth sessions from your desk. Your couch. Anywhere. http://p.sf.net/sfu/redhat-sfdev2dev ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-devel
Re: [Fink-devel] New libgettext8 package
On Wed, Dec 02, 2009 at 06:39:11PM +0100, Max Horn wrote: Am 01.12.2009 um 15:58 schrieb Jean-Fran?ois Mertens: On 01 Dec 2009, at 15:18, Charles Lepple wrote: On Tue, Dec 1, 2009 at 8:44 AM, Jean-Fran?ois Mertens j...@core.ucl.ac.be wrote: I notice there is buildconflicts for ccache-default : this breaks the operation of ccache on anything _ fink or not! _ going on in parallel.. An export CCACHE_DISABLE=1 (instead of the current env CCACHE_DISABLE=1, so as to apply also to the make command, not only the configure command) should suffice in principle ... Jean-Fran?ois, After seeing your email, I went back to re-install ccache-default, and it looks like fink did it automatically. I remember seeing the message saying fink was going to temporarily remove it, and I guess it put it back when it was done. Did ccache-default remain uninstalled on your system? No ; in general, a BuildConlicts gets reinstalled by fink at the end of the build. That's why I mentioned above things going on in parallel. Exactly. I am unhappy whenever a package forces me to uninstall ccache- default, even if it is only temporary and and it gets reinstalled later on automatically: It affects all my concurrent build activities, which is negative. Also, I wonder what happens if I try to fink build two packages *in parallel* (i.e. in different terminals) that both conflict with ccache-default , esp. if one finishes earlier than the other ? The first one's attempt to reinstall (i.e., while the second one is still building) will fail because a build_lock_ still exists, not just a fink remove this before start the process action. The whole purpose of the buildlock system is to make it safe to run multiple fink instances concurrently. Which adds to the poor user experience in this case because could not reinstall after building will cause the first fink process to fail. dan -- Daniel Macks dma...@netspace.org http://www.netspace.org/~dmacks -- Join us December 9, 2009 for the Red Hat Virtual Experience, a free event focused on virtualization and cloud computing. Attend in-depth sessions from your desk. Your couch. Anywhere. http://p.sf.net/sfu/redhat-sfdev2dev ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-devel
[Fink-devel] [Fwd: arpack package]
I'm not sure what the message meant here, since the package as sent seems to build a dylib, albeit with a compatibility_version of 0.0.0. Since arpack seems not to have been updated in a while, maybe that isn't a big problem. I'm a bit swamped right now with day job stuff, so if anybody's got some insight here, that'd be great. Original Message Subject:arpack package Date: Thu, 19 Nov 2009 19:01:00 -0500 From: Ben Abbott bpabb...@mac.com To: Alexander Hansen alexanderk.han...@gmail.com Alex, This may be no use to you, but while trying to solve problems with compiling Octave's developer's sources on 10.6 I noticed that Octave was no longer seeing arpack. When I created a dylib then Octave did see the library. Since John Eaton identified a problem in configure.in that resolved the trouble with the static version. In any event, I make an attempt to create a patch and info file. The patch does work, but after spending more than an hour on the dylib part of the info file, I'm giving up ... perhaps what I've done will be useful for you. Ben arpack.info Description: Binary data arpack.patch Description: Binary data -- Join us December 9, 2009 for the Red Hat Virtual Experience, a free event focused on virtualization and cloud computing. Attend in-depth sessions from your desk. Your couch. Anywhere. http://p.sf.net/sfu/redhat-sfdev2dev___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-devel
