[petsc-dev] broken petsc-dev related mesh?

[Jed Brown]

On Sat, Mar 12, 2011 at 19:37, Matthew Knepley <knepley at gmail.com> wrote:

> I still do not like this. I would prefer building everything in the
> directory. I do not like sticking information in the makefile.


I think the first priority should be to have all build methods (legacy,
cmake, builder.py) compile the same files. The legacy build has a technical
requirement that every file must be explicitly listed, the others have no
restriction. As long as the legacy build is supported, I think we need to
make the other builds match. There are currently some custom rules in a few
places, e.g. src/sys/viewer/impls/socket/matlab/makefile, which are only
buildable using the legacy method. The examples and tests still require it.
And both the cmake and builder stuff requires more recent Python.

At some point in the future, when all functionality is available in a
different system, we should be able to leave the legacy build behind and we
can revisit the issue of whether to list each file name or to handle
conditional compilation in some other way.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20110312/c1e0e745/attachment.html>