[ImageJ-devel] Move of the Fiji Updater into ImageJ complete

Johannes Schindelin Johannes.Schindelin at gmx.de
Tue Jul 24 21:54:03 CDT 2012


Hi all,

I tried to make as sure as I can that things do not break, but after a
frantic two weeks it is time to release the baby and give it a shot.

So what happened?

The updater moved to ImageJ2 in the same line as the launcher moved
already: the idea is to move the stuff which is not exactly specific to
life sciences to ImageJ2. Just so things are neatly separated.

The file Fiji-Updater.jar now only contains the bare minimum required to
get the ImageJ updater and run it. Probably a little less than the bare
minimum, time will tell and I will add things back.

On the good side, a ton of bugs were kicked out due to the test
infrastructure provided by ImageJ2. In fact, large parts of my debugging
were done in Eclipse with both ImageJ and Fiji updater projects loaded and
even running the classes in Fiji's custom classloader with ij.dir pointing
to a fully populated Fiji directory and setting breakpoints and single
stepping and "stuff".

So the updater should be more robust, at least in the long run.

The most notable change is that you will need to install the
ij-updater-ssh.jar from the ImageJ update site and edit the upload part of
your Fiji update site in "Manage update sites" to say (user)@fiji.sc in
the ssh host field if you want to continue to upload to our main update
site.

The other big news is that the updater can now handle versioned files. In
other words, it will consider ij-1.42a.jar and ij-1.47b.jar to be two
versions of the same component.

It still does not handle two different components requiring different
versions of a third component, though (think Micro-Manager requiring
commons-math 2.0 and TrackMate requiring version 3.0, which are mutually
incompatible).

There is still a ton of work in the updater but at least I can concentrate
on other projects for the moment without worrying about incompatibilities
betweeen the two updaters (as shown recently in the SLIM issue) or about
untested features: with the regression tests in place (which faithful
Jenkins tests continuously), I can add tests for most bugs I fix, to be
certain that I do not break things for many when I try to fix things for
one.

But for the moment I am a bit exhausted. So, please, people, try not to
have problems with the updater for the moment, m'kay?

Ciao,
Dscho



More information about the ImageJ-devel mailing list