[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Fwd: Re: init not called?]
Hi Blake,
The newest version of JANA is not backward compatible with all of
the older Hall-D code. There were some incompatible changes related to
the calling of init() routines (as some have painfully discovered). You
should be OK if you use tagged Hall-D releases with the JANA version
noted in the RELEASE notes. For example, you can see the RELEASE notes
for release-2009-02-24 here:
https://halldsvn.jlab.org/repos/tags/release-2009-02-24/RELEASE
Which states JANA 0.4.9 was used.
Admittedly, this does not make development on the trunk easy. If changes
go into one (JANA or HAll-D) that require changes in the other then one
will need to update JANA if they "update" from the repository. I think
the best way to avoid problems here is to make better announcements when
these types of changes go in. I'm guilty of not recognizing the level of
incompatibility and consequently notifying everyone of the changes
leading to this last set of problems. I'll do better in the future.
Regards,
-David
Blake Leverington wrote:
> Hi David,
>
> I see what your saying. I think my issue was just with the fact that I didn't
> know that with the newest version of the HallD software in the trunk that I
> needed the newest JANA until I started getting errors when running my
> analysis software. Will the older versions of the stable-release of HallD
> behave properly with the newer JANA? are they backwards compatible?
>
> Cheers,
> -Blake
>
>
> On Wednesday 08 April 2009 06:05:22 you wrote:
>
>> Hi Blake,
>>
>> Sorry for taking so long to get back to you.
>>
>> The short answer is yes. I personally like to keep my JANA files
>> separate from my Hall-D files just so I can easily swap JANA versions or
>> use the same JANA binaries with multiple versions of the Hall-D code.
>> Nonetheless, there should be no conflict in the naming of the files that
>> would prevent one from doing as you suggest.
>>
>> Keep in mind though that if you do change JANA versions, you will
>> almost certainly need to recompile all of the Hall-D source code. This
>> is because JANA uses templates heavily so a lot of the code is
>>
> contained
>
>> in the headers and not in libJANA.a. So in this sense, you can't have
>> one Hall-D source distribution that is simultaneously used with multiple
>> JANA versions.
>>
>> Just as a reminder, JANA is set up to act as a package much like
>> ROOT or Xerces that is installed in some arbitrary directory and one
>> sets their JANA_HOME environment variable to point to it. One just
>>
> needs
>
>> to recompile all of the Hall-D software if they decide to switch JANA
>> versions as described above.
>>
>> Regards,
>> -David
>>
>> Blake Leverington wrote:
>>
>>> Hi David,
>>>
>>> Just a comment from the perspective of a user more than a developer
>>>
> of
>
>>> the software. Is it possible to have it compile JANA in HALLD_HOME
>>>
> if
>
>>> the current version of JANA isn't up to date? I don't always have root
>>> access to the shared version of JANA on the Indiana cluster so I've
>>> had to download and compile it in HALLD_HOME anyway since this
>>>
> seems
>
>>> to have solved the errors I was getting when i was trying to
>>> recalibrate the BCAL with the updated svn trunk. I imagine other
>>> users will have to do the same.
>>>
>>> Cheers,
>>> -Blake
>>>
>>> David Lawrence wrote:
>>>
>>>> Hi All,
>>>>
>>>> Here is an exchange that took place between Matt and I earlier
>>>> today for those who are interested. The basic message is that the is
>>>> some flaky behavior in the repository trunk right now so if you are
>>>> living on the bleeding edge, be prepared to bleed.
>>>>
>>>> -David
>>>>
>>>> -------- Original Message --------
>>>> Subject: Re: init not called?
>>>> Date: Thu, 2 Apr 2009 13:06:09 -0400
>>>> From: Matthew Shepherd <mashephe@indiana.edu>
>>>> To: David Lawrence <davidl@jlab.org>
>>>> References: <AB4C8AF1-CB8A-472C-
>>>>
> BC36-91EBB37AF113@indiana.edu>
>
>>>> <49D4DBE0.4080207@jlab.org>
>>>>
>>>>
>>>>
>>>> David,
>>>>
>>>> No problem -- thanks the note -- my solution works now. If you feel
>>>> appropriate, you might send an email to the software group? I spun
>>>> my wheels for a while on this. The symptom in my case was
>>>>
> actually
>
>>>> a segfault in the event loop. I searched all over and found it was
>>>> due to a TTree Fill() command that was being called on a null
>>>> pointer (the output tree didn't get setup in init()).
>>>>
>>>> -Matt
>>>>
>>>> On Apr 2, 2009, at 11:38 AM, David Lawrence wrote:
>>>>
>>>>> Hi Matt,
>>>>>
>>>>> I've run into some problems with this too recently. In fact, I
>>>>> just traced down an issue with it last night. It has to do with
>>>>> changes in the most recent version of JANA (0.5.1). In that
>>>>> release, a somewhat significant change was made to how
>>>>> initializations were done in order to better accommodate GUIs
>>>>>
> which
>
>>>>> operate in a slightly different way.
>>>>>
>>>>> The fix seems to depend on the exact version of JANA and the
>>>>>
> Hall-
>
>>>>> D source code that is being used. For me, the main thing that was
>>>>> needed was to make the JApplication::Init() method virtual so that
>>>>> DApplication::Init() gets called properly. I'm not sure that would
>>>>> resolve the issue in your case.
>>>>>
>>>>> All I can say for now is that I'm seeing similar issues and I'm
>>>>> trying to fix them as I come across them. I'm hoping to put out a
>>>>> new JANA version (0.5.2) next week that addresses this and a new
>>>>> tagged release of the Hall-D software the week after that that
>>>>> meshes well with JANA 0.5.2. In the meantime, hopefully you can
>>>>> limp along with what you have.
>>>>>
>>>>> Sorry for the inconvenience.
>>>>>
>>>>> Regards,
>>>>> -David
>>>>>
>>>>> Matthew Shepherd wrote:
>>>>>
>>>>>> Hi David,
>>>>>>
>>>>>> I seem to have a processor that has an init routine that is not
>>>>>> getting called. I can't figure out why this is. I checked other
>>>>>> sample processors in the repository and the function seems to be
>>>>>> defined correctly:
>>>>>>
>>>>>> jerror_t init(void); ///< Called once at
>>>>>> program start.
>>>>>>
>>>>>> I got around it by calling init in the constructor, but this isn't
>>>>>> the best solution -- any ideas?
>>>>>>
>>>>>> -Matt
>>>>>>
>>>>> --
>>>>>
>>>>> -----------------------------------------------------------------------
>>>>> -
>>>>>
>>>>> David Lawrence Ph.D.
>>>>> Staff Scientist Office: (757)269-5567 [[[ [ [
>>>>> [ Jefferson Lab Pager: (757)584-5567 [
>>>>> [ [ [ [ [ http://www.jlab.org/~davidl davidl@jlab.org
>>>>> [[[ [[ [[ [[[
>>>>> -----------------------------------------------------------------------
>>>>> -
>>>>>
>
>
--
------------------------------------------------------------------------
David Lawrence Ph.D.
Staff Scientist Office: (757)269-5567 [[[ [ [ [
Jefferson Lab Pager: (757)584-5567 [ [ [ [ [ [
http://www.jlab.org/~davidl davidl@jlab.org [[[ [[ [[ [[[
------------------------------------------------------------------------