db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Rick Hillegas (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DERBY-6007) Migrate Derby website to new Apache website publication machinery
Date Wed, 19 Dec 2012 19:37:12 GMT

    [ https://issues.apache.org/jira/browse/DERBY-6007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13536348#comment-13536348

Rick Hillegas commented on DERBY-6007:

Here is what I did to implement the plan described in the previous comment:


In my website client I saved the generated content in a tarball. In retrospect, I should have
used "svn export" to do this; that would have removed the version control information in the
copy (see below).

cd .../build/site
tar cf ../../../site.tar .


In my website client:

cd .../build
svn delete site

Committed at subversion revision 1423860.


In another window and subtree, I checked out all of the content at https://svn.apache.org/repos/infra/websites/production/db/content,
added a derby subdirectory, and committed this subdirectory at subversion revision 843107.

Then in my website client, I did the following:

cd .../build
svn propset svn:externals "https://svn.apache.org/repos/infra/websites/production/db/content/derby/
site" .

svn update

Then I committed the propset in the build directory at subversion revision 1423984.


In my website client:

cd .../build/site
tar xf ../../../site.tar

Then I used the svn export command to remove the version control information which had been
tar'd up along with the directory tree. There were a number of steps here and I made some
mistakes along the way. But I think I ended up with a correctly shaped directory of files
not under version control.

I moved the contents of the doc directory to the side because I remembered from last week
that adding all of the published docs for all of the previous releases of derby took a long
time and failed at the end. I decided to checkin everything except the docs and then check
them in one release at a time.

mkdir ../z
mv docs/* ../z/
svn add *

I committed this work at subversion revision 843115.

cd docs

Then I committed the published docs, one release at a time like this:

mv ../../z/10.0 ./
svn add 10.0
svn commit ...

The published docs were committed at these revision stamps:

10.0: revision 843118
10.1: revision 843121
10.2: revision 843124
10.3: revision 843125
10.4: revision 843126
10.5: revision 843128
10.6: revision 843129
10.7: revision 843130
10.8: revision 843131
10.9: revision 843134

Then as a sanity-check, I built the Derby website in my client. A cursory inspection of the
results looked good to me. However, I noticed that the build ended up creating a file which
was no longer under subversion control:

?       build/site/broken-links.xml

I think that the "svn propset" command clobbered the hidden status of this file. I decided
to add this vacuous file to the repository and I committed that change at subversion revision
843145. Along the way I discovered the following fun fact: The re-directed generated content
subtree is not affected by commits to the larger website source tree in which it is embedded.
I think this means that when you change the website, you have to make 2 commits:

i) You have to commit the website source.

ii) And you have to commit the generated content subtree independently.

Maybe someone understands how to make this work in a more coordinated fashion.

I also discovered that if you type "svn status" high up in the website source tree, then you
will see the redirected subtree marked with an X for "externals":

X       build/site

I googled around about this and it may be that a later edition of subversion does not show
this status by default. You can suppress this status information with the following switch:

svn status --ignore-externals -q

> Migrate Derby website to new Apache website publication machinery
> -----------------------------------------------------------------
>                 Key: DERBY-6007
>                 URL: https://issues.apache.org/jira/browse/DERBY-6007
>             Project: Derby
>          Issue Type: Improvement
>          Components: Web Site
>            Reporter: Rick Hillegas
>         Attachments: derby-6007-01-aa-prepForDocsCheckin.diff
> As of January 2013, the Apache website will no longer be refreshed from the /www/db.apache.org/derby
directory tree on minotaur. Instead, projects will need to put their website content under
subversion control and tell the Infrastructure team where that subversion repository is. Most
of Derby's website is subversion controlled. However, the docs and alpha javadoc directories
are not. We will need to put docs and alpha javadoc under subversion control along with the
rest of the Derby website. Then we will need to file an issue, telling the Infrastructure
team where our content lives. For an example of such an issue, see https://issues.apache.org/jira/browse/INFRA-5215

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

View raw message