On 03/16/2016 11:05 AM, MAISONOBE Luc wrote:
Evan Ward <evan.ward@nrl.navy.mil> a écrit :
Luc,
Hi Evan,
I've been working on bug #234 where propagation crashes if two events
are close to each other. I think the fix is in
AbstractIntegrator.acceptStep() to only call reset(...) on the
currentEvent instead of all registered events. I noticed that you had
switched the behavior to reset all events in a previous commit and I
want to make sure I'm not breaking something I don't understand with
this change. All of the test cases pass in [math] and Orekit when only
the current event is reset.
I will try to have a look at this remotely. I am currently in Germany
for a conference (I just gave a talk about Orekit at ICATT and will
give a talk about Rugged in half an hour).
That's great. Hopefully there is a positive reception for Orekit and
Rugged. :)
From what you have written in the report, I think you are perfectly
right. I'll check on my side as soon as I can.
Also, would it be possible to create a quick bug fix release of Orekit
that solves this issue? I would volunteer to do it myself, but I don't
know how to perform a release.
Sure. First thing to do is to ask PMC, you can send a vote request to
all of us. If this is really urgent to you, state it and ask for a quick
response, saying that as soon as the output of the vote is known (i.e. a
sufficient number of +1 from PMC is received), you'll tally the result.
The nominal duration of a release vote is 168 hours, wich may be too
long for you.
I'll try to set up some hints on how to perform releases, but will
probably
do the work myself this time, at least for the maven artifacts. They need
to be cryptographically signed and I currently am the only one who
have the
keys (I sculd share them with other people too obviously), but I don't
know
if I can share the authentication tokens for the OSS repository at
Sonatype
to upload the signed artifacts to maven central.
Could you commit your changes as soon as possible so I can check them? If
something is wrong, we will simply create a reverting commit later on.
OK, I'll push up the fix and test case to commons-math and a test case
to Orekit. Can I start he Orekit PMC vote before I make the commit since
we vote on a feature set?