Even more webstart codesigning changes

[manics]

The commit logs should explain everything, main points:

• If http_proxy and/or https_proxy environment variables are defined they will be converted into the form required by Java when signing with timestamping
• Removed release-webstart, add release-webstart-unsigned (no signing) and release-webstart-signed (self signed). release-all calls release-webstart-unsigned
• Removed all existing jarsigning code, everything is done using omero_insight_sign.py (so update-version won't sign, since there's no way for it to know whether it should or not)

Possibly for discussion:

• omero_insight_sign.py sits under docs/hudson but is called from build.xml
• docs/hudson/OMERO.sh calls release-all to created unsigned artifacts, then signs the zip with omero_insight_sign.py. I could add a release-all-signed target instead to sign the jars in dist/lib/insight before zipping but this would mean we're not following the release signing workflow so closely.

Since most of the original PRs were done in a rush this should probably have a proper review- I've created a dummy PR at manics#2 which I think includes all changes (#2069 #2088 #2193 #2241 ... unless I've missed one).

[joshmoore]

@manics: major thanks for bearing with us for all these changes, but I have to ask: what's the definitive place to look for how to use all these wonders you've built?

[manics]

docs/hudson/omero_insight_sign.py will print out a help message. There were some changes to the Inisght manifest in an earlier PR (which is why we can't just sign an existing 4.4.10). Everything else is basically making sure things aren't self-signed for release, without breaking the expectations of developers who want things self-signed for non-release builds. It just took a few iterations and dead-ends to get there.

[sbesson]

Failed http://ci.openmicroscopy.org/job/OMERO-5.0-merge-daily/626/console. Excluding.

[manics]

This is due to different jarsigner versions. I added a check to see if jarsigner was executable, on Java 1.7 running it without arguments prints out a help message and returns 0, on gretzky (1.6) it returns 1. Thoughts? Take out the check, or allow 0|1 as a return code and hope that works on all versions/platforms?

[kennethgillen]

86e8891 of omero_insight_sign.py will sign jars.

[joshmoore]

(ome1)jamoore@blue:/opt/ome3$ git show simon/webstart_codesigning-dev_4_4:docs/hudson/omero_insight_sign.py > /tmp/sign44.py
(ome1)jamoore@blue:/opt/ome3$ git show simon/more_webstart_codesigning_tweaks:docs/hudson/omero_insight_sign.py > /tmp/sign50.py
(ome1)jamoore@blue:/opt/ome3$ diff /tmp/sign*

i.e. the just merged 4.4.11 version is identical to this new tweaked version, so in my opinion, 🚢. Will do tomorrow start-of-day unless there are objections.

[joshmoore]

Missed start-of-day, but merging regardless. Thanks ever so much, @manics

[manics]

--rebased-to #2321