Re: compilation failure workaround (RH 5.2 / 2.22.10)

From: Matthew D. Langston (langston@SLAC.stanford.edu)
Date: Mon Aug 09 1999 - 08:10:44 MEST


Hi Rene,

Rene Brun wrote:
> 
> We are maintaining binaries to simplify the export procedure.

I don't understand.  Wouldn't it be even more simple to export only
source code?  You would, of course, still want to insure that ROOT
passes a comprehensive testsuite on those platforms on which ROOT is
known to work before making a public release.  However, actually
maintaining binary versions for all platforms on which ROOT is known to
work seems like a nightmare.

I understand that you have users who actually need for you to provide
them with a binary ROOT distribution (e.g. Windows users and specific
collaborations), but it seems that there might be tremendous value in
emphasizing to everyone else that ROOT is primarily distributed in
source form, and that it is the recommended way of downloading and
installing ROOT.

Once the ROOT CVS repository is in place, you could then perhaps ask the
community for volunteers, one for each supported platform, who would be
responsible for sending the ROOT team testsuite results just before
releases.  Distributing responsibilities such as these out into the
community might dramatically whittle down the time that the ROOT Team
spends on platforms which aren't their primary development platforms.

The CVS repository would probably also encourage more frequent bug
reports for those users who track, or even use, the development branch
of the repository.

> Just to give an idea: in the past two years we had more than 40000
> downloads of the binaries compared to 8000 for the source.  We know
> that we have to improve the makefile mechanism.  This is becoming
> pretty complex, in particular with the many dialects of Linux,
> compiler versions, glibc, etc.  I was hoping more stability in this
> area.

I would offer again that GNU Autoconf, Automake and Libtool solve *all*
of ROOT's configuration and portability problems.  These packages are
well tested and have active maintainers (one of which is even a member
of roottalk).

This work was already done for ROOT once, and could be easily integrated
into the current version of ROOT.

The only disadvantage is that build times are about 80% longer.  This
seems like such a small price to ask ROOT end-users (note that I did not
say ROOT developers) to pay in order to have a highly portable ROOT
which builds identically on all platforms, from Unix to Windows to ....

ROOT developers (e.g. the ROOT Team) would not have to pay this 80%
price in build times (unless they just wanted to).  Only end-users would
have slightly longer build times, but they only have to build ROOT once,
not several times like developers do.

> Concerning CVS, Fons has nearly finished the procedure to
> create/maintain the ROOT CVS site.  He will make an announcement once
> this is ready.

It is good to hear that Fons has nearly finished the conversion to CVS.
Perhaps it would be a good idea to announce to roottalk the design of
the CVS repository in order to get feedback from other CVS users before
it is completed.  Fons may appreciate the collective wisdom available
from the community, as I know that there are several long-time CVS users
who subscribe to roottalk.  There might be some real gems of wisdom out
there.

--
Matthew D. Langston
SLD, Stanford Linear Accelerator Center
langston@SLAC.Stanford.EDU



This archive was generated by hypermail 2b29 : Tue Jan 04 2000 - 00:43:37 MET