From: Gordon L. Kindlmann <glk@uc...> - 2012-12-20 15:49:17
I've released Teem version 1.11.0 this morning (roughly in keeping with the predicted Dec 19 release date):
The text of the release note is viewable there. The main change I made since the previous emails indicating the planned release was the following:
BUG FIX (or rather a 99% fix):
ten/experSpec.c/tenDWMRIKeyValueFromExperSpecSet() had been saving DWI
gradient directions to KVPs in the NRRD header with only "%g %g %g"
instead of "%.17g %.17g %.17g" (the analog of which has been used for
the rest of the floating-point nrrd meta-data for quite awhile). This
caused needless significant precision loss in how the gradients were
saved. However, there is a still a subtle (almost negligible) change,
caused by the normalization done at the end of
ten/chan.c/tenDWMRIKeyValueParse(). The problem is that the computed
norm of what should be a unit length vector is not *exactly*
unit-length, so the normalization, which should sometimes be a no-op,
actually changes the values very slightly. Fixing this will require
more thought or an API change.
This means that NRRD DWI headers should be more accurate now as written records of DWI experiments created within Teem, though there are still some subtle issues to resolve.
In the end I didn't finish third derivatives in gage, or the gageMulti infrastructure, but I did get the tests passing at least on the computers I have access to (including on big endian machine). I would have liked to fix these PNG warnings:
but it seems like they've been fixed multiple times before and at this point I've lost track of what the right thing to do is. What the heck?
In the trunk I've already started indicating that new commits will be towards a 1.11.1 path release, which will hopefully be in early January.