Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Migrated to Confluence 4.0

ReleaseManager Update

Navid has prepared a rebuild of ST 09-15-05 on Mac tiger using ROOT v5.20.00-gl5 and python 2.5.1-gl4.  Richard had reported troubles after installing via Tony's web installer.  Basically the python installed hung with no output even with the verbose flag on.  Ultimately the trouble turned out to be a zero-length python binary, where it is believed the web installer does not handle hard links succesfully.

Next up are RHEL5-64 ST builds after the u35 outage we will have our first trial build.

Developer releases are being created but we need an updated RMViewer to gain access.  About 5 minutes of work.

In an effort to fully suport VC90, we are considering the new SCons checkpoint release 1.2.0.d20090919 but there are some issues on Windows concerning the bindexplib workaround.  Navid plans to look at this after addressing ST RHEL5-64 builds and copying over OBF for RHEL4-32.  This newer SCons also requires a modification to msvs.py in the site_tools directory.  Either msvs.py will be SCons version specific (only on windows) or Joanne will find a way to introduce a conditional statement.  In the meantime, CVS will not be updated until Navid has a change to work out the issue with bindexplib.

ASP Update

ASP-to-SCons+conversion+status+by+package

Externals

Not sure about FSW status on 64 bit machines should talk to Tony Waite.

Heather is working to set up CLHEP 1.9.2.2 on VC90 which is useful for both Gaudi and G4.

OBF RHEL4-32 needs testing and is now available

Tracy will get back to G4 on VC9 when able

Branch Tag Question

We had discussed the new branch tag convention and we settled on:   <packageName>-xx-yy-zz-ab##

Branch names something like GR-branch-xx-yy-zz 

RHEL5-64 builds require package owner help for unit test problems.

RMViewer RC2 is now available with developer releases.

New SCons checkpoint release 1.2.0.d20090919

ASP Update

https://confluence.slac.stanford.edu/display/SAS/ASP-to-SCons+conversion+status+by+package

Externals

Setting up a meeting concerning FSW status on 64 bit machines with Tony W, Tracy, Jana, and Gregg.

OBF RHEL4-32 needs testing and is now available

Need to try out G4 builds on RHEL4-64 and RHEL5-64

Schedules

Image Added

Minutes

ST builds on RHEL5-64

Some bugs in SConscript files have been fixed.  What remains are failures in the unit tests (similar errors also appear for RHEL4-64) due to floating point errors or integer errors where there are assumptions concerning 32 bit code.  Need to contact package owners to work out fixes.

RMViewer and Developer Release downloads

Since the meeting, Navid has looked into the odd behavior Heather saw when GLAST_EXT is set to an area with no write-access,and there are missing externals. Apparently, this could produce a race condition, and Navid has fixed it. Patch will be available in an upcoming RC3 version of RMViewer.

Preparing for checkpoint release of SCons 1.2.0.dwhatever

Joanne will be looking to see what might be different about her installation that may be preventing the newer version of SCons from working.  Meanwhile, we need to check that we can maintain both VC71 and VC90 on the same build machines where we will be running CMT and SCons builds.  We chose to install VC7.1 on glast05, so Navid could try out a CMT build and make sure having VC90 available does not cause CMT trouble.

Intro of RHEL5 machines into batch

Richard mentioned the introduction of RHEL5 machines into the batch farm and was wondering if we have build machines identified.  Navid stated bldlnx14 and 15 are RHEL4-32.  bldlnx10 is RHEL5-32.

End of RHEL3

Heather wondered when we will stop the RHEL3 builds.  Probably when SLAC drops the remaining RHEL3 boxes.  RIchard suggested we should send email to the collaboration as a heads up - that has since been done.

ASP

Joanne is just about ready to pass something to Jim for testing.

Externals Organization on Windows

From Joanne via email:

On the V disk under the GLAST_EXT directory there are 4 subdirectories:
VC8
VC8debug
Windows-i386-32bit-vc71
Windows-i386-32bit-vc90
The first two are for CMT, the second two for SCons.  But, because of the new-style path for SCons which includes a compiler component, we don't need those 2 top directories; we could get by with one, as I do on my local copy of externals.  And there is a significant advantage for those who go back and forth: you can use the same definition of GLAST_EXT in both cases.
So I propose we add a new directory under GLAST_EXT, say for the sake of discussion it's called Windows-SCons (maybe 32bit should be in the name somewhere?) and the structure underneath for a typical external looks like this:
Windows-SCons
             |
            ape
              |
...|--------------------| -----------|
 2.4                  2.5.1          2.5.3
   |                         |                |
|-------|         |-------|        |---------| 
vc71   vc90    vc71   vc90    vc71       vc90
At some point we could get rid of the old directories Windows-i386-32bit-vcXX  but only after RM knows about the new structure.
Probably we can't do anything similar for SLAC Linux because there are too many varieties and compiler name isn't sufficient to distinguish.
This discussion is ongoing, as Navid recalls that we left this directory structure alone so as to avoid disrupting the builds and installer.

Schedules

Navid wondered how we could all update the list of activities and dates.  Heather has since created a Google Doc for that purpose, which allows editing:

http://spreadsheets.google.com/ccc?key=0AradgsCW6fMzdDhiUklXdW84bkI3Wk5DX2JwWU9uTVE&hl=enHeather wanted to be clear what our guidance would be concerning the initial branch tag, whether we would suggest GlastRelease-01-02-03-gr00 or gr01.  We prefer gr01.  Doc in the workbook will be updated to reflect that.