On Friday, January 07, 2011 10:58 AM, Noel O'Boyle wrote:
>Thanks for the patches, John,
You're welcome.
>I the Python patch with one modification; if PYTHON_PREFIX is
>specified then DESTDIR is ignored. I assume that this is fine with the
>package maintainers.
You will do as you think best, but I don't like that change, and I don't think
it's wise. These two variables have distinct, non-overlapping purposes. For
example, if there were ever a need to put an RPATH into a binary component of
the Python extension, or a hardcoded path into the Python module, then you
would want it to include the PYTHON_PREFIX, if any, but not the DESTDIR.
Furthermore, making an exception for that case leaves the OpenBabel support for
DESTDIR inconsistent with convention under some circumstances, and it requires
more work in some cases similar to mine by requiring that the same thing needs
be specified twice (at least).
>I'm still not happy with supporting DESTDIR though (and wish that
>CMake had not decided to do so). I would prefer adding staging support
>through a CMake configuration variable such as -DDESTDIR=.
That could have worked, but I am glad that CMake adopted the existing
convention instead of inventing its own.
Note also that DESTDIR has other uses than just staged installations. For
example, if you maintain a chroot environment then you can use DESTDIR to
install to your chroot without building directly there. If you have multiple
chroots then with DESTDIR you can install to each of them without reconfiguring
or rebuilding anything. Alternatively, if you want to check what's going to be
installed before you actually install it, then DESTDIR is just the thing for
you, and an install-time decision is exactly what you want.
> The main
>problem with adding support for a build-time variable, is that we
>really need to apply conditional logic at install time, but this does
>not seem to be possible with CMake (correct me if wrong).
I'm still not much of a CMake guy, and in my limited research I did not find
any easy ways to persuade CMake to perform logic at build time. I suspect
there are some, however, because CMake Makefiles call back to CMake. I did
have to overcome a similar problem, though: to get the DESTDIR variable to be
evaluated and interpolated at build time instead of at configuration time. You
will see that I managed that by using \$ENV{DESTDIR} in CMakeLists.txt instead
of just $ENV{DESTDIR}; at configuration time the former is converted to the
latter, and at build time the latter is interpreted as the value of DESTDIR.
I'm not seeing a need here to apply conditional logic at build time, however.
The standard DESTDIR paradigm is that you just prepend $(DESTDIR) to all the
installation paths in your Makefile. If DESTDIR is given during "make install"
then the installation paths automatically reflect the specified alternative
root; if not, then $(DESTDIR) expands to nothing, and files are installed
directly to their final places. Why should the build system otherwise care or
do anything differently depending on whether DESTDIR is given?
Cheers,
John
Email Disclaimer: www.stjude.org/emaildisclaimer
------------------------------------------------------------------------------
Gaining the trust of online customers is vital for the success of any company
that requires sensitive data to be transmitted over the Web. Learn how to
best implement a security strategy that keeps consumers' information secure
and instills the confidence they need to proceed with transactions.
http://p.sf.net/sfu/oracle-sfdevnl
_______________________________________________
OpenBabel-Devel mailing list
OpenBabel-Devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/openbabel-devel
