We unanimously agreed to this, and we decided to take this even further to generally discourage use of %define in favour of %global. I created a proposal for the guidelines at

http://fedoraproject.org/wiki/PackagingDrafts/global_preferred_over_define

It also contains the Packaging/* pages that need editing, which I eye-ball-checked that they only contain trivial %defines.

On Tue, Apr 24, 2007 at 05:33:27PM +0200, Axel Thimm wrote:

On Mon, Apr 23, 2007 at 12:11:35PM +0200, Axel Thimm wrote:

...

Or to place it more explicit make

%{!?python_sitelib: %define python_sitelib %(%{__python} -c "from distutils.sysconfig import get_python_lib; print get_python_lib()")} %{!?python_sitearch: %define python_sitearch %(%{__python} -c "from distutils.sysconfig import get_python_lib; print get_python_lib(1)")} [%{!?foo: %define foo ...}]

to

%{!?python_sitelib:%global python_sitelib %(%{__python} -c "from distutils.sysconfig import get_python_lib; print get_python_lib()")} %{!?python_sitearch:%global python_sitearch %(%{__python} -c "from distutils.sysconfig import get_python_lib; print get_python_lib(1)")} [%{!?foo:%global foo ...}]

Otherwise using parametrized macros can mess things up (#237448). The nesting levels seems to get mixed up and the conditionalized definition above get swallowed away. The %global ensures that this bug doesn't happen.

OK, looks like the bug is indeed in using this construct with %define inside. rpm's lazy garbage collection pollutes the global macor namespace until a more involved nesting construct appears which properly cleans up.

So we really need to change %defines within a %{...} to %globals.

Let's vote on this today unless s/o thinks he needs to investigate further.

...

It may also be cleaner to remove the space between the colon and the percent sign.

On Tue, 2007-04-24 at 21:46 +0200, Axel Thimm wrote:

We unanimously agreed to this, and we decided to take this even further to generally discourage use of %define in favour of %global. I created a proposal for the guidelines at

http://fedoraproject.org/wiki/PackagingDrafts/global_preferred_over_define

I agree with this in the case of large macro expansions or cases where nesting might occur. Incidentally, there's a very limited stack size on macro expansions (8192 bytes currently) that is an annoyance when you've got deeply nested macros. I'd love to see that limitation go away :-)

Jon.