Maintenance

Created a script in svn/hcmc/exist which cleans the git tree completely, builds exist, signs the jars, creates a fairly minimal dist, and builds a test instance with one of our own xars, then starts it for testing. All that remains to be done is figure out how to change the jetty.xml file and any oother conf files to add the required placeholders.

Still on the quest of the best way to build a working eXist and update it cleanly, we've been testing the dist-zip build target. First, though, I determined how to build with DS's patch for the bug I reported the other day, which is still only a pull request (#1256); CLI for future reference:

git pull
git checkout 1256
git fetch origin pull/1256/head:1256
git checkout 1256

Then we do:

./build.sh clean
./build.sh 
./build.sh jnlp-all

the last of which signs the jars, enabling the Java client to work. This gives us a working build in the repo context, but of course it's cluttered up with lots of unwanted stuff, including source. That's why we wanted to use the dist-zip target, but the result is currently broken due to an obscure timestamping issue which I reported today on the gmane list. We could manually create a clean tree, of course, but it would be easier to have a script do it.

GL's desktop was rebuilt, and the huge TCCD git repo was copied over. However, it had been checked out using git://, and somehow it knew that the old desktop had SSH keys configured on GitHub, so we had to generate new keys (GL couldn't remember the pw for the old keys), add them to ssh-agent, and add the public key to GitHub (the key must be for the email address by which GitHub knows you). Then pulls and pushes would work. All this is quicker than checking out that repo again, of course.

Helped GN set up the new TV, and got the condenser mic working for video-conferencing. The TV's own speakers are sufficient, probably, but I've connected the others anyway.

URLs containing "editor" are all being redirected to the HCMC site, even though only the www/editor URLs were supposed to be; that borks our adaptiveDB projects, so I've written to pts to see if we can get it fixed. MS's two projects Bilibin and St Pete were also screwed up, being in a "projects" folder, but in that case, I've just moved the projects up one level to the hcmc/www, and we'll cope with the changed URLs.

Since it's not trivial to get this done, these are the steps:

• Ask NETS to add the domain to the UVic DNS servers, pointing at the root of hcmc.uvic.ca.
• Ask sysadmin to set up the apache virtual host on HCMC's web cluster.
• Test locally.
• Ask the person with the domain record to log into their registrar and set the DNS IPs to UVic's servers.

For the record, this one points at /home1t/thewali/www/

Janelle
153 Research
154 Admin unused Research

Judy
47 Research
48 Admin (phone)

Stewart
44 Research
43 Admin (phone)

Martin
55 Research
57 Research
58 Admin (phone)

Pat

233 Research
234 Research
56 Admin (phone)

Greg

59 Research
60 Research
195 Admin (phone)

55 Research
57 Research
58 Admin (phone)

<?xml version="1.0" encoding="UTF-8"?>
<project basedir="." default="all" name="test">
    
<!-- This is a test file for figuring out how best to 
     integrate file diffing checks into ant build files. -->
    
<!-- We need ant-contrib to do foreach.   -->
    <taskdef resource="net/sf/antcontrib/antcontrib.properties"/>
    
    
<!--    Two filesets to diff. -->
    <fileset id="files1" dir="${basedir}/files1" includes="*.txt"/>
    
    <fileset id="files2" dir="${basedir}/files2" includes="*.txt"/>
    
    <target name="diffOneFile">
        <propertyregex property="otherFile" input="${inFile}" regexp="files1" replace="files2" global="true"/>
        <echo message="diffing ${inFile} against ${otherFile}"/>
        <trycatch>
            <try>
                <fail message="${inFile} is different from ${otherFile}.">    
                    <condition>
                        <not>
                            <filesmatch file1="${inFile}" file2="${otherFile}"/>
                        </not>
                    </condition>
                </fail>
            </try>
            <catch>
                <echo message="ERROR: DIFF FAILURE..."/>
                <exec dir="." executable="diff">
                    <arg line="-u"/>
                    <arg file="${inFile}"/>
                    <arg file="${otherFile}"/>
                </exec>
                <fail message="Build failed because of differences between  ${inFile} and ${otherFile}. See diff output above."/>
            </catch>
        </trycatch>
        
    </target>
    
    <target name="diffFileSets">
        <foreach target="diffOneFile" param="inFile">
            <path>
                <fileset refid="files1"/>                
            </path>
        </foreach>
    </target>
    
    <target name="all">
        <antcall target="diffFileSets"/>
    </target>
    
</project>

At HC's request, created a TEIC org and a jenkins repo on hub.docker.com, and pushed the image up there. I first had to do a clean build after deleting every docker image that could be found on the system (docker rmi -f tei-jenkins && docker rmi $(docker images -a -q)), then I built (docker build -t teic:jenkins .) and tagged the build (docker tag 61c5096c86b2 teic/jenkins) then pushed it (docker push teic/jenkins). Took a while to figure out how to do that, so I'm documenting here. I think pushing an image only works when it's tagged with exactly the right org/repo tag.

Spent much of the day working on this bug, which although it afflicts our TEI plugin, actually seems to be caused as far as I can see by a Saxon bug; but there's something about running Saxon in the Oxygen context which makes it happen, while running exactly the same transformation command outside of Oxygen does not fail. No solution yet; we've rolled back the plugin release for the moment.