On 2006-11-11 16:13-0800 Alan W. Irwin wrote:
The two remaining things I need to do are to make a script that does what I
did by hand to create the tarball and also make some CBS changes so that
prebuilt files in the tarball will be used in the build when they are
available.
I am now done with an initial version of a script called make_tarball.sh
(that's a reuse of the filename of an old script of mine to create the
distribution tarball which has not been used in a long time.)
This new script adapts make-cvs-tarball.sh to our CBS build system. In
addition, I have made some simplifications (for example, I have dropped all
the fooling around with version so all the version changes must be committed
to CVS before invoking this script). So run
./make-cvs-tarball.sh -h
first to see what is possible. I have worked hard on that
self-documentation so that is probably all you need. N.B. don't expect
everything to be identical to make-cvs-tarball.sh. (When in doubt, I have
dropped some of the finesse of make-cvs-tarball.sh to create something more
straightforward/easy to understand.)
I have tested this script with the following invocation:
[EMAIL PROTECTED] ./make_tarball.sh \
-c -i /home/software/plplot_cvs/installcmake -o -DENABLE_java=OFF \
make_tarball.sh_out
The options concern some tests you can do with the created tarball.
-c says unpack the created tarball and build it with cmake and make
and test that result with ctest.
-i /home/software/plplot_cvs/installcmake specifies the install prefix
for the tarball test configure and runs make install after cmake, make,
and ctest.
-o cmake options allows you to specify extra cmake options for the
tarball test configuration. In my case, I must disable java
because of java bugs in Debian stable.
I have also tested the generated tarball by hand using the deprecated ABS
./configure; make; make check commands, and all seemed to be well.
Orion, I suggest you generate a tarball with this script, and then check
whether that tarball is ready to create an rpm for your RedHat distro
using our CBS.
Release status:
One nagging issue still outstanding is the run-time trouble for Tk-related
applications that Hazen discovered on Mac OS X. It would be nice to get
this fixed, but I don't think it is release critical (so long as we mention
it in the release notes) since this issue has probably always been with us.
Also, I have not yet made the changes so that our CBS system can take
advantage of prebuilt files in the tarball, but again I don't believe this
is release critical (and I may get to it before the release in any case).
My conclusion is our source tree is now ready for a development release
featuring our new CBS.
Hazen, some changes obviously need to be made to the release process to
avoid using our ABS (and thus, to test that our CBS is up to the task of
making a release). So if you ever find yourself ready to invoke either
cf/bootstrap.sh (except inside the above script to prepare files that are
going into the tarball) or ./configure during the release process, don't!
Instead, use the CBS alternative. For example, in the above script you will
see the following:
cmake -DCMAKE_VERBOSE_MAKEFILE=ON \
-DPREBUILD_DIST=ON ${DOC_ARG} \
-DWWW_USER=${WWW_USER} ../plplot cmake.out
followed later by
make prebuild_dist make_prebuild_dist.out
DOC_ARG defaults to -DBUILD_DOC=ON by default and -DWWW_USER=${WWW_USER}
transforms to -DWWW_USER=hbabcock by default. The make prebuild_dist
command builds the documentation and copies it back to the source tree
before the subsequent make package_source creates the source distribution
tarball with the appropriate cpack command.
Afterward, to install the identical documentation that was built for the
tarball to the website, you merely have to do the following commands by
hand:
cd /tmp/plplot-dist-prep/build_dir/doc/docbook/src
make www-install
Hazen, I think the timing of the next release (5.7.0) is now in your hands.
Obviously, you will need some extra time beyond what you normally spend on a
release to figure out and document the changed CBS-only procedure for
release that I have alluded to above. In particular,
README.Release_Manager_Cookbook is now pretty dis-jointed with add-ons of
everybody's different comments, and this would be an excellent opportunity
to turn that into a straightforward cookbook of how you prepare the release
with our CBS.
Let us know when/what you plan to do. Also, contact me off list if you run
into any release generation issues for our CBS.
Alan
__
Alan W. Irwin
Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).
Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the Yorick front-end to PLplot (yplot.sf.net); the
Loads of Linux Links project
