April Projects
21-April-2010
- Continue to fix SL5 compilation and configuration issues as they
are discovered. Expect this to take a dozen hours or so the first week,
dropping off substantially thereafter.
- We've got claims that the warning list is intolerable. Do we want
to start dealing with the warnings before they turn into errors? Or
issue a hack to turn them off?
- Get development into proper SRT release shape. Lynn says from 2-14
days, as most of the work is already done. Other demands on her time
may dominate.
- What impact will this have on development-lite?
Review how this is used.
At Wisconsin I followed the directions: copied the files from
the web page, executed them, and edited the crontab file for daily updates:
00 04 * * * ${HOME}/execute_rebuild 1> ${HOME}/er.log 2>&1
00 05 * * * rm ~/.k5login
Changes to development-lite appear automatically and that pesky .k5login
file (which interferes with ssh in our cluster) gets removed. So far
no problems.
- Get 7.0.1 into new SRT framework and fix compilation issues. Lynn
says from 2-7 days. I found a few odd compilation problems, but
these are probably known already. (I pulled it to Wisconsin and tried to
compile there.)
- 7.0.1 release seems to be defective: include file not in place,
missing link, etc. On SL5:
TrackingObjects/Storable/CT_HitSet.hh needed a
#include "CotGeometry/CT_Calibration.hh"
SvxDaqMods includes a reference to V00-00-83, which doesn't exist in this
release.
The #define HepStd std doesn't seem to be propagating in
CLHEP using rootcint
We need geant/v3_21_14-GCC3_4_3/Linux+2.6 but it
isn't there. (fixed 26-Apr)
There are some assert complaints, which were probably already solved
by Lynn.
There's a stack size too small with java in COTCalibDBTables
There's a segmentation fault ?!? in compiling
CalorCalibDBTables/TdcPhaSlewing.java
- Compilation/link 7.0.1.migrate under SL4 and SL5
- 10K Validation: Link/Run SL3/SL4 vs SL5/SL5, then SL4/SL4, then SL4/SL5
Probably takes a week (clock time) after port shakedown if no problems
- Read/write cross-platform tests. Can be done
in parallel with the 10K tests.
- Iterate several times, check for readability, contents
- Stress test: get close to maximum file size.
- If it fails this 7.0.1 becomes an experts-only "non-release"
- Plan for May projects
List and who
- Fixing configuration issues: Lynn
- Development into SRT release form: Lynn
- Compilation fixes in development: Lynn says she is best to do this,
but I see a lot of things with her name on them already.
- Development-lite discussion: All. I thought this was not going to be
a problem, but perhaps I'm wrong
- Products for development; check and recommend: ?
- 7.0.1 configuration repair and getting this into SRT release form: Lynn
- Compilation fixes in 7.0.1 on SL5: JNB?
- Products for 7.0.1; check if OK for SL4/5: ?
- 10K validation for 7.0.1: ?
- root cross-platform read/write and stress tests: ?
Notes
development should probably turn into development.migrate for testing, then
we backport everything to development and use that.
We agree that development-lite is not broken and won't break and remote
sites that don't do a cron update ought to.
radon and bld4_nfs distribute SL3 and SL4 respectively.
cdfcode and bld4 build SL3 and SL4 respectively
bld5 builds and distributes SL5--though the latter is not fully tested
because it is also a build machine
By August at the latest all farm machines should be SL5
General plan by most to keep building SL3 development until August
I'm not persuaded that this is necessary--I thought SOP was to pull
the code from the head and build locally, not pull libraries. Can we
find out?
We agree that 7.0.1 only requires 10K for our .m release, for the
physics testing. I still say we need a more thorough shakedown for
root interoperability--provided this has not already been done. But
nobody has told me yet where this was tested.
cdfcode also has
- code browser: Donatella and Lammel Is sam browser fixed?
Donatella says the code browser and the postgres DB and the dedicated
web service are all on cdfcode. It may be possible to add non-cdfsoft
directories, but 11-May is earliest she can work on this.
- Some distribution tar files. Lynn to move them. Needs to have
a different home than the build??
- Web pages somewhere?
- Database for code browser?
- Database for what else?
- sam update (for sam browser, which doesn't work, so why is this running?)
- update_dcaf--what is this for?
postgres /usr/bin/postmaster -p 5432 -D /var/lib/pgsql/data
postgres postgres: stats buffer process
postgres postgres: stats collector process
cdfsoft /bin/sh /home/cdfsoft/pull_dbstats
cdfsoft tee /home/cdfsoft/maint/log/dbstats/work.log
cdfsoft /usr/sbin/sendmail -FCronDaemon -i -odi -oem -oi -t
cdfsoft /usr/krb5/bin/rcp -X cdfdb fcdfweb.fnal.gov /cdf/work/mysqldata/remoteLogServer/archive/dbsmonsum_mysql_20100426.tar.gz /cdf/spool/cdfsoft/Backups/dbstats/dbsmonsum_mysql_20100426.tar.gz
Search and see
April Decisions
21-April-2010
- When do we stop building on SL3?
- After validation of 7.0.1
- Why do we have nightly builds? If the release was frozen, shouldn't
it only need a rebuild-check if something has changed in the OS?
- Which SL3 hardware needs to be returned, which discarded, and which
can be reloaded?
- Do we need new machines for the ILP? Guessing we can continue with
the current non-SL3 build machines.
- Do we argue for a new release? And if so, based on 6.1 or 7.0? (I would,
and barring surprises, based on 7.0.1)