SCons 0.96.90 (testing pre-release for 0.97) is available

We're doing this because there have been a lot of changes to the SCons
infrastructure since 0.96.1 was released, and we want some extra assurance
that we've caught as many bugs as possible that may have slipped through
any cracks in our regression test suite. Here's how you can help:

BUGS:

    If you run into any outright bugs, please file a report at our
    SourceForge bug tracker:

        http://sourceforge.net/tracker/?group_id=30337&atid=398971

    Also please send some email to either dev@scons.tigris.org or
    user@tigris.org.  We'll be monitoring the mailing lists for bug
    reports, but the bug tracker report will help make sure that
    these things get tracked and, we hope, fixed.

    WHEN YOU REPORT A BUG, please be specific.  It's more effective
    if you don't try to guess at the cause, but instead describe the
    symptom you observed *and* the differing behavior you expected.
    Cutting and pasting is good.

    If possible, it's a *huge* help if you can reduce your
    configuration to a small/simple test case that clearly
    demonstrates the problem.

RECOMPILATIONS ON UPGRADE:

    In general, we try to make sure that upgrading SCons releases
    doesn't cause recompilations.  But with the significant changes
    in this release, we know there are configurations that will
    experience recompilations.  Unfortunately, we don't have a concise
    definition of what configurations get recompiled on upgrade.
    You can help us track this down by running with the following
    option the *first* time you run SCons 0.96.90 against your old
    build tree:

            scons --debug=explain

    If you experience recompilations, please collect the messages
    explaining why the target(s) are being recompiled and send them to
    dev@scons.tigris.org.  (No need to file a bug report this time.)

PERFORMANCE:

    The most complicated (and highly-anticipated) changes in this
    release involved speeding up SCons and reducing how much memory
    it uses.  To set expectations clearly:  don't expect these changes
    to have suddenly made SCons as fast as Make; internally, SCons
    is doing a lot more than Make does!  Instead, our tests suggest
    that a typical small-to-medium configuration can probably expect
    a 10%-30% improvement from the changes in this release.

    However, these changes should significantly speed up large
    SCons configurations, configurations with long CPPPATH lists,
    or that were seemingly "hung" due to inappropriate scanning of
    large directory trees.  In other words, this release of SCons
    should be significantly more scalable than previous releases.

    IF THIS RELEASE PERFORMS WORSE FOR YOU:  We really really really
    want to know!  Changes that help one configuration may actually
    hurt others, and we want to know what these trade-offs are so
    we can make some conscious decisions about how to handle them.

    IF THE PERFORMANCE IMPROVED SOME, BUT NOT ENOUGH:  We want to
    know that, too.

    TO REPORT PERFORMANCE ISSUES, please run SCons on your
    configuration with the following options enabled:

            scons --debug=memory --debug=count --debug=memoizer

    Then please send email to dev@scons.tigris.org and file a bug
    report at the tracker (URL above), attaching a copy of your
    configuration and a copy of the debug output from the end of your
    run.  Note that the --debug= options above will make SCons
    take a little longer to build.

    If you can, it would also be *extremely* helpful to run SCons
    on your configuration separately with the following option:

            scons --profile=scons.prof

    This will take a *lot* longer to build (probably an order of
    magnitude) but it will dump Python profiling information to
    the "scons.prof" file.  Sending us the file will provide a
    very detailed picture of where the bottlenecks are in your
    configuration.

Thank you for your help and support in improving this release!