[ImageJ-devel] Legacy Mode work

[Johannes Schindelin]

Hi,

On Wed, 19 Dec 2012, Johannes Schindelin wrote:

> On Wed, 19 Dec 2012, Curtis Rueden wrote:
> 
> > Thanks for the update; that is great news. Barry & I did some testing,
> > and found a few issues, but given the time constraints we are going to
> > go ahead with the beta6 release this week.
> > 
> > One big issue I found is that ImageDisplays that originate in IJ2 land
> > (e.g., from File > Open) do not get translated to IJ1 mode unless you
> > execute a legacy command on them first. That is, more generally, the
> > command for switching to legacy mode apparently does not explicitly
> > invoke any IJ2<->IJ1 synchronization, meaning they can be out of sync
> > with ImageJ1.
> 
> Good point, I forgot to test that.
> 
> It is probably some addition to LegacyImageMap#toggleLegacyMode(boolean)
> to make sure all known ImagePlus instances are shown.
> 
> Will have a look during the flight.

I did work on that. The only problem I had during testing was that at some
stage, the x and y axes would not be found, so the PixelProbe threw
exceptions abound. After a while, that was fixed, though. Probably some
performance issue preventing things from being harmonized quickly enough.

BTW I just saw that Jenkins is not happy, it thinks that legacy tests are
failing. But when I ran mvn locally ("mvn clean test"), things worked!

> > I took a look at adding some sort of legacyService.sync() call, but
> > the LegacyService API still needs work. Currently there is a lot of
> > logic taking place in LegacyCommand.LegacyCommandThread.run(), which
> > should ultimately get factored out to be more accessible from the
> > LegacyService itself. But that work will have to wait until after the
> > holidays.
> 
> Agree!
> 
> > So anyway, we will note this limitation in the release announcement,
> > for now.
> > 
> > Another problem which I would like to see fixed before release is that
> > on OS X, when starting up ImageJ2 now, there is a VerifyError with
> > class MacAdapter that was not previously present. I hope to track this
> > down tomorrow morning and fix before cutting the release.
> 
> I think I know what it is. Left over from some previous iteration.

Yep, it was that issue: I added special handling for "$service" in the
CodeHacker so that I could easily change the method how the LegacyService
is discovered for use in the javassisted functions, but I forgot one.

FWIW I pushed my stuff to the 'legacy' branch, but the commits were
designed to be patched into the branch ;-) I'll work a bit more on the
commit messages.

For the moment, I will have to refrain from working on the Jenkins issue
and from trying to resolve my ImgLib issues because my internet connection
is too slow... :-(

Ciao,
Dscho