Getting to the finish line with SCons, RM, and GoGui

Minutes

Attendees:  Joanne Bogart, Emmanuel Cephas, Richard Dubois, Navid Golpayegani, Heather Kelly, Karen Heidenreich

SCons Status
Navid reports that the ScienceTools are essentially ready.  One issue is that as ST packages are updated, package owners are not keeping their SConscript files up to date.  Navid is doing his best to keep those SConscript files updated.

Emmanuel reports that there are about 20 packages to go for GlastRelease SCons builds.  The remaining issue blocking completion are two externals:  OmniOrb and OBF.  Emmanuel is working on OmniOrb 4.1.2 now.  As for OBF, we desire to reorganize to fit the new directory structure to be used by SCons.  Obviously, Tracy will need to be included in this discussion.  Joanne pointed out, that at least on Linux, symbolic links could be utilized to provide a structure more convenient for SCons, and also maintain the structure closely mimicing that used by FSW. Unfortunately, symbolic links are not possible on Windows.  Later in the meeting , we further discussed OBF, and agreed that Navid and Emmanuel should go ahead an reorganize OBF as they see fit to get the GR SCons builds going.

We're currently using SCons version 1.0.1

Richard asks what version of GR are used for the builds.  Navid replied:  GR HEADs

Joanne requests that the ScienceTools externals.scons be updated to use the current version of ROOT v5.18.00c.  Navid will do that.

Heather asked Emmanuel that the 20 packages left to do for GR be documented in Confluence so we can keep track of status.

RHEL5 status: Navid is getting the various externals prepared, hopefully ready by tomorrow.

64 bit - Navid has all the externals, just needs to update MySQL

Mac - we're strictly interested for ST for now - do we need to grab the externals from Eric?

Windows  What we have now in SCons will fully compile our code.  The additional directives to actually build the MSVC++ project files needs to be added into our setup.  Navid notes that Tracy maintains the SCons produced project files are not useful in the debugger.  Perhaps we can plan for a demo during core week.  Before then, we can send around what we have to Tracy and Toby and start to get feedback.

 Release Manager

 Now uses Qt, and is up and running.  There is no front end yet.  This RM is transitioned to use the new MySQL maintained by SCS, not the one on glastlnx01.  The table structures have changed.  Richard asked if the new tables are documented?  Not yet, but they will be by tomorrow.  Using that information, Karen will hopefully be able to come up with a schedule to rework the RM web front-end.  In fact, Navid has such a page now:

http://confluence.slac.stanford.edu/display/SAS/Release+Manager
Next up is the TagCollector rewrite, should be able to reuse some code from the version written a couple of months ago Navid was wondering if it should be a separate tool or part of GoGui.  Richard wondered why is the TagCollector a separate web tool from the RM pages?  Navid reminded us that the old TagCollectgor was located on glastlnx01, but started up some services on glastlnx02.  There were too many levels of indirections, causing it to break very often.  Now we are looking for a single application tha handles the CVS calls.  Richard also asked whether adding tags would be a priviledged function?  Yes.  Reading would be open, but creating tags would require authenication.  Richard asked if we should use Tom's scheme?  Need to discuss that with Tony, concerning line mode authenication.

Concerning the rewrite of the web front end..  Richard asked if we could hand draw the desired look of the pages?  Much of what exists can transfer over, though there are design changes from CMT to SCons.  Navid offered to draw up he has in mind.  Be helpful to get Tony (and Joanne) to take a look and offer suggestions.  Providing a list of all the functions we want, such as the summary pages.  A more complete list, will allow Karen to offer a more realistic sense for how to scope out the job.  Of course, whatever we come up with, will be iterated on.

GoGui

Joanne pointed us to the "To Do" section of her page:http://confluence.slac.stanford.edu/display/SAS/GoGui+Status+and+Laundry+List

debug requires SCons build something like CMT setup, something like the wrapper scripts

Some items need input from Navid, such as overriding a package for developing

tag and rtag function (like MRvcmt)

Joanne plans to install GoGui on central linux

How to get to the Finish Line

Show CMT and SCons builds are equivalent.

OBF needs to be repackaged to complete GR builds.  Navid has a script to extract the FSW includes and libraries.

Navid asked when we could move to the new tagging convention.  The consensus is that we need to provide tagging utilities first, through GoGui and the tag collector.
Joanne asked about handling branch tags for packages other than checkout packages.  We agreed on the suggestion: xx-xx-xx-grx

Is it ok to start JO transition?  (Navid)  Recall Navid received permission some weeks ago, but has not had a chance to start using symlink tricks to migrate the JO files into a top-level JO directory for the SCons builds.  We agreed that Navid could carry on, but should just send a note to corelist mentioning that he has done so.

Heather plans to migrate the CMT externals to the same directory structure used in SCons.

How does the move to SCons affect the installer? (Richard)  Navid responded - we need a new one.  This time around we will have one java installer with a command line option.

Planning for a test version of SCons/RM/GoGui in time for the core meeting.

Navid and Emmanuel will maintain status on Confluence pages for RM and SCons builds, and Joanne will carry on with her GoGui page.

 Agenda

Release Manager

  • Current status of new RM - Navid
  • Status of RM Web Pages - Karen 
  • Tag Collector

SCons

  • Current status of SCons GR builds - Emmanuel
  • What externals are needed to complete GR SCons builds?
  • What version of SCons are we currently using?
  • Windows support

GoGui

Externals

  • Migrating externals used in CMT to same directory structure as SCons
  • No labels