Core Minutes 25 November 2014

Core Minutes 11/25/2014

Present: Joe Asercion, Joanne Bogart, Jim Chiang, Warren Focke, Heather Kelly, David Landriu, Leon Rochester,

FSSC: (Joe) They're planning on a new round in February, currently working on the Yosemite release. For that they needed to update Python.

There is an ongoing discussion on how best to deal with Python updates. Heather has been leaning towards Anaconda [as neatly defined by Joe: "a closed source python distribution that includes a separate repository optimized for data analysis"] in order to reduce the hand labor involved in making a release. FSSC is leary of incorporating such a large, complicated externally-maintained object. We may just have to agree to disagree.
Pass8: (Leon) Recent news includes work on understanding the isotropic background, various kinds of mopping-up, and a bug fix for the crash Johan encountered when looking into the redhat6 64-bit of the newest GR.

(Joanne) and JJ tracked it down. We discovered that the event being passed to obf code was not consistent with itself: although it claimed to be of a particular length, in fact the contributions from TEMs, ACD and GEM took up much less space; the event was padded with 0's out to the advertised length. Apparently when the false length differed from the proper length by too much, obf would get confused and crash (exactly why this causes a crash is not yet clear, but the events certainly were malformed). We then turned our attention to EbfWriter, which creates the event and ultimately found where the size calculation goes awry: a place where the code was using sizeof(pointer-variable) where it should have been sizeof(object-being-pointed-to), namely an int. On a 32-bit machine they're the same; on a 64-bit machine they're not. The fix has been committed and tagged. In the limited testing I did, it appears to take care of the problem.
Core talk:
- Mac (Joanne) No progress since last meeting since all my Fermi cycles were devoted to the obf crash. Maybe next time.
- New fsw & GR (Heather) All available evidence indicates we don't need a new GR to go with the new fsw release. It does not affect obf. It does not affect science data. The one thing it does affect is the ordering for tracker calibration data, but we don't access that data directly, only via metadata (which is cognizant of the change).

minutes index

J. Bogart, Last Modified: