How to Configure Sherlock
*************************

1. What can be configured
~~~~~~~~~~~~~~~~~~~~~~~~~
There are three different levels of configuring/customizing Sherlock:

  - runtime configuration in cf/sherlock and cf/local (see doc/config)

  - compile-time configuration: config switches set before compiling, selecting
    optional features (see sherlock/default.cfg)

  - customization: a set of C declarations and functions, which define word
    types, attributes and the ways to match them. In most cases, you should
    use the "free" customization designed for searching in web pages and similar
    documents. If you want just a simple search engine for indexing databases,
    the "customs/bare" customization can be better. If you want to extend Sherlock
    beyond that, you will probably need to write your own customization -- just
    follow the structure of the "free" customization and the comments there.


2. How to configure
~~~~~~~~~~~~~~~~~~~
To set up compilation for a given customization, possibly overriding its default
compile-time options, just run

	./configure <customization> [<option> | -<option> | <option>=<value> ...]

The default values of feature options are taken from sherlock/default.cfg and modified
by <customization>/config. Compiler flags and options dependent on compiler, OS and
CPU type are set in lib/autoconf.cfg. Everything can be overriden by options specified
on the configure's command line, which have the highest priority.

If you want to see the resulting set of options, take a look at obj/config.mk.

Options specifying compiler/linker/debugger options can be also overriden during
compilation by `make <option>=<value>'. While it's also possible to specify the other
options in this way, it probably won't have the desired effect, because configure
also generates C include files containing the options.


3. Where to build
~~~~~~~~~~~~~~~~~
If you run configure in the source directory, it prepares for compilation inside
the source tree. In this case, an `obj' subdirectory is created to hold all generated
files (object files, binaries, generated source files etc.) and all final files
are linked to the `run' subdirectory. No other parts of the source tree are written into.

Alternatively, you can compile in a separate object tree (which is useful when you
want to build several different configurations from a single source tree). In order
to do that, switch to the destination directory and issue `<source-dir>/configure ...'.
This way, configure will create the `obj' and `run' directories locally and set up
a Makefile which refers to the original source tree.