14-Sep-1998

Hello,

	This is a brief summary of some of the progress we have made in
the topics: CVS, Makefiles, Unix Platform support. This is not the
formal write-up expected from PasRec*, has not been reviewed by the
parties involved, and should be seen as merely a draft e-mail to let
people know ASAP what approximately has happened and/or been agreed upon.
I expect corrections and more detail will follow.

	I, Rob Harris, and Rene Brun have met to discuss these issues.
As agreed, we for the most part stuck to the CDF-Unix side of issues,
though there is large overlap with CD/D0-Unix issues too. The only
sub-topics I missed I think were the issue of root compatibility with
C++ exceptions (used in D0 code), thread safety (on-line and data
browser folks have some interest), and HP support (some D0
collaborators may want this). Gordon Watts will presumably take these
up, along with NT issues, with Rene Brun on Thursday/Friday this week.

	Some makefiles of interest to us already exist in Root v2.00/11
(Linux-kai, SGI-kai), but they have some minor problems. Only the
Linux-Kai one builds the core of root correctly at this time, and can
only build working executables with Kai C++ v3.3. SGI-kai fails while
building a dictionary with both v3.2 and v3.3, which is also seen in
Linux-kai but for some reason does not prevent a build. The makefiles
do not yet build h2root, which I think is crucial to the distribution.
I am installing makefile support for h2root today. I will document the
remaining issues and send them to Fons for consideration.

	On the CVS front, we agreed to put root into the Zoom
repository (as SoftRelTools is, with no implication of Zoom support) in
such as way that local folks can easily check it out and build it to
suit. The CVS repository will be fed by the cmz "repository" through
the source code distribution of root. Each root release would be put
into the Zoom repository en mass, not as diffs. FNAL will feed defect
and modication requests back to the root team, using the local
repository to test suggested changes. FNAL-specific changes, for
instance to makefiles, will be re-applied if necessary if they are not
accepted by the root team in time for a new release. Some minor
re-organization of the product will take place (move headers all to
"include", create source sub-directories), but these issues are not
high priority. Root is open to this level of re-organization as part of
the path to their using CVS to support root. That switch by the root
team would not happen in the short-term, however. Rob Harris agreed to
put root into the Zoom repository once I have a version ready to be
used in this context, probably on Tuesday 15-Sep.

	Root would like to drop some binary distributions platforms,
maintaining  support for source code distribution only. An example
given is IRIX6-Kai, where there has been nearly zero interest in binary
distribution off the ROOT webpages. Robert and I agreed with this since
we prefer to have a local build to support CDF and D0's tastes in C++
options, and we would like to move to newer compiler versions faster
than Root.

	CDF has agreed to make an account available on cdfpca (as on
cdfsga), and later on b0du02 (Dec Unix), for the Root team. I just need
to locate some disk space and generate the official account request
forms. I pointed out to them that even if they do not use these
platforms for development and testing, they can use the systems to
document the official FNAL OS/compiler environment on each. They have
not ported to the Dec Unix - Kai C++ combination yet, so some root
development will need to take place before that platform is fully
supported. Future platforms, such as Sun/Solaris-Kai are of interest to
the root team too (CDF is considering a proposal to support
Sun/Solaris). The root team has had difficulties with shared libraries
under AIX.

	Rene and I also agreed on a general approach for joint
investigation and/or development on test suites and bug tracking
systems. In the short-term, the current approach will continue to be
used to avoid slowing the development effort. Testing will continue to
be done with a growing suite of test programs, examples, and benchmarks
to which FNAL can contribute. Bug tracking will continue to be done in
e-mail with an e-mail archive. Meanwhile, FNAL and root can investigate
the available options for testing and defect tracking, several of which
are freely available. Rene agrees that the current approach for testing
and defect tracking will not scale as root's functionality continues to
expand, as significantly more folks use root, and as the list of
platforms supported expands. The time scale for adaptation of root
development to a large-scale testing and bug tracking system would be
no later than when root goes into "managed co-development" by FNAL and
the root team, perhaps in 6-12 months.

Rob K.