After a (much-too-long) hiatus, I just pushed a batch of changes to my
notmuch repository.

This includes the several changes that I recently reviewed on the
mailing list and said that I had committed locally and would push soon.

Additionally, I've pushed a few of bug fixes as follows:

  * Fix to never add a partially complete message to the database.

    Various interruptions to notmuch processing, (perhaps other bugs
    described below) would leave a half-complete message in the
    database. And later runs of notmuch would trip and fall over on this
    half-complete message. Thanks to a suggestion from Olly, notmuch now
    never adds the partial documents to the database so this problem
    should be avoided entirely.

  * Fix for really long message ID values.

    Scott mentioned various Xapian exceptions and notmuch internal
    errors that occurred after trying to add a message with a 300+
    character message ID. I've fixed this to now be handled well, (with
    new tests in the test suite to handle 512-character message IDs).

  * Make things easier after "make install" (for Linux at least)

    On Linux systems, the configure script will now identify the various
    paths supported by ldconfig (such as /usr/local/lib). Then, if the
    prefix is to one of these places (such as /usr/local/), make install
    will run ldconfig automatically. This will allow some people to be
    able to more easily run notmuch immediately after installing.

    Similarly, if the notmuch library is installed to some non-system
    directory, (such as --prefix=${HOME}/foo), then the final link of
    the shared binary will include a DT_RUNPATH entry to the directory
    in which the library is installed. This will also allow the notmuch
    binary to work immediately after "make install" without the user
    needing to set the LD_LIBRARY_PATH variable.

I implemented this last piece only on Linux. I think OS X doesn't need
additional support for this since linking binaries against libraries
might automatically encode the paths, (though I haven't tested). I also
haven't investigated what might be the right thing to do for Solaris,
(it's probably not all that different than what I did for Linux).

So if anybody wants to put together similar patches as needed for their
favorite platforms, then that would be great.

-Carl

-- 
carl.d.worth@intel.com