RE: Tigris binary packages for Windows
Hi David, list, On 2010-04-22 17:06, Cooke, Mark wrote: I am resurrecting this thread to ask if anyone has come forward to volunteer time and/or effort to resurrect the windoze binaries as we are still on 1.6.6 against 1.6.11 announced a few days ago. From: David Darj [mailto:z...@alagazam.net] Sent: 22 April 2010 21:21 I have built both 1.6.9 and 1.6.11 They are available on my webpage http://alagazam.net You (and anyone else) is welcome to download and use it. The reason I've not announced the release in this (users) list is that I've hoped some people reading the dev list (where I did announce it) to download and test it first so I know my build environment is okey. As the web page says all test on subversion itself is running ok., but the bindings has not been tested. On 2010-04-23 10:29, Cooke, Mark wrote: You star! Thanks very much. I will do some testing with apache 2.2 on windoze using mod_dav_svn and python bindings (to Trac using mod_wsgi) and report back... From: David Darj [mailto:z...@alagazam.net] Sent: 23 April 2010 18:08 Thanx that would be gr8. So far so good. Minor points but the .11 package readme still refers to .9 and some of the files have a different name format from the tigris packages, e.g.: o svn-python-1.6.6.win32-py2.6.exe o svn-win32-1.6.11_py.zip ...although I realise one is an installer and the other just the files. Actually, I am unsure about installing the latter, can I just copy the files over the ones I already have installed or do I need to tell python about it somehow (probably OT for this list I guess, perhaps Troy is listening)? Out of interest, did you use VC6 or VC2008 to compile (there were suggestions earlier in the thread that there might be some potential issues using VC2008 binaries against the official apache build which I am required to use here): It's built using the good old VC6 like D.J Heap previously did. That way no runtime-dll:s (msvcr*) needs to be distributed with it. Excellent, thanks again. Also, did you hook up with Troy about the installers ~ I assume that he would be able to put them on tigris at least until we get a new home on apache: I havn't had any contact with troy, but this mail is cc:d to him as well. So if he is reading it it would be great if he (or anyone else) want to put the there. It would be really great if Troy could pick up the baton here as that would negate my python install issues. All the best, ~ mark c
externals update fails if repository copied
I have svn copied a repository folder to a second location, and then tried to reference some of the (new folder) contents at a fixed revision in an svn:externals. svn update does not find the revision, however, an un-revisioned reference works. Moving the revision reference to the copy-creation point works, but I maintain the externals with a script which finds it hard to know at what revision an object was created. svn info returns the last changed revision as the ancestor-last-changed, which my script looks at. Anyone any ideas ? -- David Kitchen InSync Technology Ltd 13B, Dragon Street Petersfield Hampshire GU31 4JN +44 (0) 1730 235 117
svnserve uses lots of memory (after blame?)
We have a repository with 5000+ revision and the db directory size is about 1.2GB After doing a blame on a file with a long history the svnserve memory usage is going up to 850MB and stays there. I'm not sure what happens after a long running time, bug after doing a full checkout on a restarted process the memory consumption is only 30MB. Is that a normal caching behavior that can be controlled or a bug ? We are using version 1.6.6 on a windows server. Regards Danielle
Re: svnadmin hotcopy fail with 1.6.11
On Apr 25, 2010, at 2:05 AM, Stefan Sperling wrote: On Sat, Apr 24, 2010 at 10:15:47PM -0700, Andy Peters wrote: Hello, list ... I just updated a server from 1.6.3 to 1.6.11, running on Fedora 7. I didn't do a dump and load on the repository. When running svnadmin hotcopy on the repo, I get a complaint about file fsfs.conf not found. When I looked in the repo, I didn't see this file, so that explains the error. But what has changed between 1.6.3 and 1.6.11 that would cause this error, and better yet, how do I fix it? There was a bug in versions prior to 1.6.11 which caused fsfs.conf not to be copied during hotcopy (it had simply been forgotten). From http://svn.apache.org/repos/asf/subversion/trunk/CHANGES : Version 1.6.11 (19 Apr 2010, from /branches/1.6.x) http://svn.apache.org/repos/asf/subversion/tags/1.6.11 User-visible changes: [...] * make 'svnadmin hotcopy' copy the fsfs config file (r905303) Since you hotcopied your repository with versions of Subversion which had the bug, you lost your fsfs.conf (probably some time ago). Now svnadmin complains if the file isn't there. But you can easily recreate it. Either copy it from a freshly created repository, or create an empty file called fsfs.conf (either way will give you the default configuration). I'd rather not make this error non-fatal because people might be relying on fsfs.conf and not realise that it has disappeared. Stefan Perfect! I found some examples online and it seems like the options offered by fsfs.conf don't apply to me, so I simply created an empty file and that all works now. Thanks, -a
Re: svnadmin hotcopy fail with 1.6.11
... spoke too soon. svn doesn't like an empty fsfs.conf file; it complains about missing section headers when trying to do an update or whatever. I copied an example from somewhere and all's well. -a
Re: Tigris binary packages for Windows
On 2010-04-26 13:58, Cooke, Mark wrote: Hi David, list, On 2010-04-22 17:06, Cooke, Mark wrote: I am resurrecting this thread to ask if anyone has come forward to volunteer time and/or effort to resurrect the windoze binaries as we are still on 1.6.6 against 1.6.11 announced a few days ago. From: David Darj [mailto:z...@alagazam.net] Sent: 22 April 2010 21:21 I have built both 1.6.9 and 1.6.11 They are available on my webpage http://alagazam.net You (and anyone else) is welcome to download and use it. The reason I've not announced the release in this (users) list is that I've hoped some people reading the dev list (where I did announce it) to download and test it first so I know my build environment is okey. As the web page says all test on subversion itself is running ok., but the bindings has not been tested. On 2010-04-23 10:29, Cooke, Mark wrote: You star! Thanks very much. I will do some testing with apache 2.2 on windoze using mod_dav_svn and python bindings (to Trac using mod_wsgi) and report back... From: David Darj [mailto:z...@alagazam.net] Sent: 23 April 2010 18:08 Thanx that would be gr8. So far so good. Minor points but the .11 package readme still refers to .9 Oh...thats my fault missing to change the version number in some spots. I think I was focused on getting the dependecies versions right. and some of the files have a different name format from the tigris packages, e.g.: o svn-python-1.6.6.win32-py2.6.exe o svn-win32-1.6.11_py.zip ...although I realise one is an installer and the other just the files. You're right, the first one is an installer and the second one only the files. If you look at tigris download page a bit further down you see the corresponding zip-file there as well. I've just havn't looked into byinding these installers. Actually, I am unsure about installing the latter, can I just copy the files over the ones I already have installed or do I need to tell python about it somehow (probably OT for this list I guess, perhaps Troy is listening)? I'm not a python programmer so I don't really know how to install this package (thats why the bindings are not tested). But I suppose you can just extract the file over the old ones. Actually it's not Troy that used to build the python installers, it was DJ Heap, who also used to build the binaries. Maybe he has some info on this matter. Out of interest, did you use VC6 or VC2008 to compile (there were suggestions earlier in the thread that there might be some potential issues using VC2008 binaries against the official apache build which I am required to use here): It's built using the good old VC6 like D.J Heap previously did. That way no runtime-dll:s (msvcr*) needs to be distributed with it. Excellent, thanks again. Also, did you hook up with Troy about the installers ~ I assume that he would be able to put them on tigris at least until we get a new home on apache: I havn't had any contact with troy, but this mail is cc:d to him as well. So if he is reading it it would be great if he (or anyone else) want to put the there. It would be really great if Troy could pick up the baton here as that would negate my python install issues. All the best, ~ mark c
Re: Subversion update bug ?
Mon, 26 Apr 2010 16:34:28 -0400, /Phil Pinkerton/: svn 1.6.5 while in a working copy I modified a file for testing. After testing I wanted to update the file I modified in the working copy to the version in the repository. the resulting update did not replace the files in the working copy as expected the modified file still remains. the only way I could have the update work was to delete the modified file then do the update. Why did svn not update the modified file ? Svn has not updated the file because there's probably nothing to update it with. If someone has committed a new revision of it in the meantime, svn would update the file trying to merge both - your changes and the new changes from the repository. What you really want is to revert the file: http://svnbook.red-bean.com/en/1.5/svn.tour.cycle.html#svn.tour.cycle.resolve.revert http://svnbook.red-bean.com/en/1.5/svn.ref.svn.c.revert.html -- Stanimir
Re: Subversion update bug ?
On Mon, Apr 26, 2010 at 16:34, Phil Pinkerton pcpinker...@gmail.com wrote: svn 1.6.5 while in a working copy I modified a file for testing. After testing I wanted to update the file I modified in the working copy to the version in the repository. the resulting update did not replace the files in the working copy as expected the modified file still remains. the only way I could have the update work was to delete the modified file then do the update. Why did svn not update the modified file ? It may have and you didn't notice. Changes made since your last checkout/update are merged into your local copy silently (well, not 100% silent, you should get a merged indication). Any conflicting changes (you and the other user modified the same line, but in different ways) should flag a conflict.
Re: svnadmin hotcopy fail with 1.6.11
On Apr 26, 2010, at 12:40 PM, Daniel Shahaf wrote: Andy Peters wrote on Mon, 26 Apr 2010 at 11:28 -0700: ... spoke too soon. svn doesn't like an empty fsfs.conf file; it complains about missing section headers when trying to do an update or whatever. I can't reproduce this with trunk or 1.6.x. I suppose I should be clearer. Without fsfs.conf, the Subversion client complained. I copied an example from somewhere and all's well. You could have run 'svnadmin create' and copied the fsfs.conf file of the resulting repository. (Same for ~/.subversion/config files, by the way.) Ah, good to know. I just did as you suggested. I don't know if it's better but it certainly can't hurt. -a
Re: svnadmin hotcopy fail with 1.6.11
Andy Peters wrote on Mon, 26 Apr 2010 at 15:59 -0700: On Apr 26, 2010, at 12:40 PM, Daniel Shahaf wrote: Andy Peters wrote on Mon, 26 Apr 2010 at 11:28 -0700: ... spoke too soon. svn doesn't like an empty fsfs.conf file; it complains about missing section headers when trying to do an update or whatever. I can't reproduce this with trunk or 1.6.x. I suppose I should be clearer. Without fsfs.conf, the Subversion client complained. I do not get a complaint with an empty or non-existent fsfs.conf: % cd /tmp % svnadmin create r % svn co -q file:///tmp/r wc1 % svn up -q wc1 % rm -f r/db/fsfs.conf % svn up -q wc1 % touch r/db/fsfs.conf % svn up -q wc1 % (same output with 1.6.x and trunk)
Re: Can I have one project link to two subversion server?
Thanks for all reply. I like the subversion since it is a nice tool running both on window and linux. I looked git and svk and it seems they are only running on linux and I am uisng windowXP. I really hope that svk can become more stable and mature and runs on window. Anyway, here is reason I want to check in the code into a private resository before I merge the code into the office SVN. During the development, I added a lot of codes for debugging in many stages. I want to keep track these changes during the development, so if anything does not work, then I can easily trace back to find a working version. All these debugged codes are not supposed to be checked in to the official SVN. After all debugging steps, I will clean the code and test it. Once it is working, then the final code will be checked into the official SVN. By doing this, I can add debug code, print out the results and will not worry that later I lost track if something I added break the code. Thanks Frank On Fri, Apr 23, 2010 at 4:19 PM, Curley, John john.cur...@windriver.com wrote: Hi all, Sounds to me like the developers want to avoid branching and merging. Try it. You might like it. Seriously though, if you have concerns with branching and merging, I suggest some training and make some trial runs in a test repository. I believe there are software tools that will help you manage the branching and merging. HTH, John