[Halld-offline] Offline Software Meeting Minutes, June 10, 2015

[Will Levine]

Hi all,
A few words about git.

I'm not involved in GlueX software development anymore, so this is not
meant to be a yea/nay vote on the issue of git for GlueX, but just me
sharing my experiences with git to help you all make that decision.

I think Mark's slides de-emphasize the thing about git which I think
will have the greatest benefit for GlueX offliners: the ability to
easily create and switch between local branches. There are a couple of
scenarios where this is really useful.

1) Before checking in your changes, you want to test your new code vs
the old code. The way I did this with svn was making a backup copy of
every file I modified, then using svn revert on the modified files,
doing the tests with the old code, then copying the modified files
back to their original location. This is a really annoying procedure
and it's easy to lose your work if you're not careful. With git what
you would do is commit your changes to a local branch, called say
bcal_fix, then you can switch back and forth between old and new
versions by saying `git checkout master` or `git checkout bcal_fix`.
It's really easy, and it's impossible to lose your work.

2) You are working on two different projects, say a change to HDDM and
a change to TOF reconstruction, that are totally independent and
neither is ready to be checked in. You should be able to work on these
two projects independently, test them separately, so that if your HDDM
work breaks something temporarily that doesn't stop you from working
on your other project. This really difficult to do with svn, but
trivially easy with git. You just have two branches 'hddm_work' and
'tof_work' and when you want to switch projects you just commit your
changes to the current branch and checkout the other branch. This is a
very common use case and git handles it so much better.

The downside of git is that it can be can complex and confusing (I
like the two diagrams in this blog post comparing svn workflow to git
workflow: http://stevebennett.me/2012/02/24/10-things-i-hate-about-git/).
The man pages are comprehensive, but cryptic. My feeling is that for
people who just need to make an occasional bugfix, git will be a pain,
but it will make things easier for heavy developers.

Will

On Thu, Jun 11, 2015 at 10:18 AM, Mark Ito <marki at jlab.org> wrote:
> Folks,
>
> Please find the minutes of yesterday's meeting below and at
>
>
> https://halldweb.jlab.org/wiki/index.php/GlueX_Offline_Meeting,_June_10,_2015
>
>   -- Mark
>   ________________________________________________________
>
> GlueX Offline Meeting, June 10, 2015, Minutes
>
> Present:
>
> CMU: Curtis Meyer
>
> JLab: Mark Ito (chair), David Lawrence, Kei Moriya, Eric Pooser, Simon
> Taylor, Beni Zihlmann
>
> NU: Sean Dobbs
>
> Announcements
>
> New Releases. Mark reminded us of the recent releases of HDDS and Sim-Recon
> as announced on the email list.
>
> Jana 0.7.3: David has release a new version of Jana. Details will be coming
> out shortly but some highlights are:
>
> event numbers changed from int32 to int64 to accommodate long runs
>
> implementation accommodates 32-bit event numbers currently used by sim-recon
>
> search for associated objects done recursively, the new default, depth of
> search can be limited by user
>
> a list of all possible command line options can be generated
>
> Spring 2015 Commissioning Simulations
>
> Sean, Justin and Mark have put in a lot of adjustments to the conditions for
> the simulation, mainly to correct oversights. There is now a web page
> listing these conditions in addition to Sean's original wiki page.The system
> is ready to go, but we will wait a bit for new version of the CDC track
> finding software from Simon before we start. For now we plan not to use the
> new BCAL code from Tegan Beattie while it goes through validity checks by
> the Calorimenter Group.
>
> Offline Monitoring
>
> Kei summarized the state of the periodic offline monitoring jobs. See his
> slides for the complete story. He covered:
>
> statistics for the current launch (last Friday)
>
> a problem with the farm last weekend that delayed processing
>
> the status of the start counter plug-in
>
> deprecation of the 2-track skim
>
> success/failure comparisons between the launches for the Spring data
>
> a problem with FCAL ADC data
>
> crashes due to an EVIO parse error
>
> schedule for the next launch (June 19)
>
> David will send out an email describing how to avoid the EVIO parse error.
>
> Updating Information from the Analysis Workshop
>
> Sean reminded us that part of the guidance for new collaborators, students
> in particular, is to review the analysis workshop we held in the Summer of
> 2013. He reported that the details of some of those activities have changed
> significantly since then, a cause of confusion. Since the concrete examples
> are stored in a virtual machine (which avoided the problem of everyone
> having to build their own system) they are not easy to update.
>
> Eric told us that he has a wiki page with notes on the sessions and
> solutions to the problems. He will link it to the workshop wiki page.
>
> We discussed ideas to address the problem but did not come to firm
> conclusions. We might want to hold another such workshop at some point.
>
> Git for GlueX?
>
> Mark gave us his take on issues surrounding a possible transition from
> Subversion to Git for a source code version control system. Please see his
> slides to see all the bullets (they are also on the DocDB). He discussed:
>
> Conversion of reconstruction code (i. e., not everything under Subversion)
>
> Multiple repositories in Git
>
> Git commits: staging code before checking it in
>
> Some common Git commands
>
> Workflow models: how to organize the multiple repositories
>
> Nice Git features
>
> Extra steps in implementing and using Git
>
> How to manage Conversion?
>
> Why we should change
>
> We did not make any decisions; in fact we did not event discuss a
> decision-making process. There are some things to work out before a change
> can be proposed, mainly how the group would transition to a new system
> without seriously impeding code development work.
>
> Retrieved from
> "https://halldweb.jlab.org/wiki/index.php?title=GlueX_Offline_Meeting,_June_10,_2015"
>
>
> _______________________________________________
> Halld-offline mailing list
> Halld-offline at jlab.org
> https://mailman.jlab.org/mailman/listinfo/halld-offline