Importing TWiki installation.

Jeroen Coumans jeroen at linuxfromscratch.org
Sat Jul 17 05:37:47 PDT 2004


Jeremy Huntwork said the following on 16-07-2004 21:06:
> On Fri, 2004-07-16 at 13:10 -0400, Jeremy Huntwork wrote:
> 
>>Well you can obviously see I'm learning by trial and error. :)
>>Jeroen, I ran into a bit of trouble importing into the branch you made,
>>test.lfs.org, so I made a new branch, not realizing that I was dumping
>>more info in than I needed.  If it is usable for our needs, great,
>>otherwise we can delete the whole branch.  I'll be testing out post-
>>commit actions to a temp directory shortly.

Ok, I'll clean the regular branches up when I do the repository 
reorganization.

> A few notes on updating the site via post-commit:
> It seems the easiest way to do it is checkout the branch into the web
> directory, nicely updating only the files that have changed.  So the web
> directory needs to be a svn working directory with a .svn folder.

Yep; an update is also the fastest. We can filter out the .svn 
directories with .htaccess. And for rsync, we should also filter it out.

> The user running Apache("httpd"?) is the one that calls the hook script
> post-commit, so all files checked out (in theory) will be owned by that
> user and group, as will the .svn directory.

Apache will need read-access to the website repository, otherwise 
someone will need to check it out as httpd & use his credentials for 
caching.

> (BTW, I'm only finding info on repository-level hook scripts. Does
> anyone know if similar scripts can be made/used for branches?)

A branch is just a directory. You can use "svnlook dirs-changed" to see 
which one was changed (only on the server). But it's probably better if 
we determine via properties which branch should check out/update where 
(eg. property www -> www.lfs.org will get updated).

> So it seems the easiest way to
> turn /home/httpd/test.linuxfromscratch.org into an svn working directory
> is to copy an existing .svn directory there and "chown -R httpd:
> httpd .svn"

No; see above. httpd needs read access; this is the safest method. The 
post-commit will need to be run as httpd:httpd and this should check out 
the appropriate directory or update it if it already exists.

> Once that is done we can add a line to post-commit that checks out the
> branch.
> If we setup the site as stated above, someone with appropriate
> permissions needs to set that up. :)

To be specific, we need write access to the repository "hooks" directory.

> Comments?

Yes, last one: post-commit is silent by nature. In order to see it's 
output, I use the following as post-commit:

#!/bin/sh
#   [1] REPOS-PATH   (the path to this repository)
#   [2] REV          (the number of the revision just committed)

/usr/local/bin/post-commit "$1" "$2" >> 
/var/log/subversion/post-commit-${2}.log 2>&1

And "set -x" in /usr/local/bin/post-commit.

-- 
Groeten/Greetings,
Jeroen Coumans
{faq,website}@linuxfromscratch.org
www.jeroencoumans.nl



More information about the website mailing list