Loadsa patches...

[Andrew <an...@aaaaaaaa.demon.co.uk>]

David J. MacKenzie <dj...@va.pubnix.com> has sent us a large number
of fixes for various parts of the distribution  including modifications
to documentation, configuration files and support programs.
Unfortunately these have been provided in just 2 mail messages so
it's going to take a while for me to break these up into Apache-standard
patch formats.  I've said as much to David so perhaps he'll help
by resubmitting them in a more usable form.  Here are David's
comments.  Give me a shout if there's something here that catches
your eye and I'll try and extract just those relevent bits of
patchery straight away.

David:
* Makes query and post-query not allocate gobs more memory than
  they will ever need.  We run CGI programs here with a soft limit of 1
  megabyte of data space, and without this patch, these trivial programs
  exceed that limit!  That isn't great for performance on other systems,
  either, though it's not likely that they're run often.
* Removes an unused variable and fixes another one's type in util.c.
* Adds some missing MIME types.
* Makes the new dbmmanage print a useful usage message.
* Makes httpd_monitor more portable, and useful in places with more
  than one web server.
* Fix typos in the API doc.
* Add missing prototypes.  We like to compile things with
  gcc -Wall -Wstrict-prototypes to catch possible errors.
  foo() is a K&R declaration that doesn't say anything about its
  arguments, whereas foo(void) says foo takes no arguments.
* Allow the password and group files in .htaccess files to be relative
  pathnames (to the server root), to make moving or mirroring a server
  less painful.
* Don't print errno in error messages; print the human readable string.
* Try to improve an error message that has confused dozens of CGI
  program developers in my experience, who don't understand that a
  "malformed header from script" may well mean that their CGI program
  didn't run, or core dumped, or wasn't written correctly and didn't even
  attempt to write out an HTTP header.  It would be even better if Apache
  logged the signal that killed the CGI program if it died that way.
  I haven't taken the time to figure out how to do that; I have done
  it on NCSA 1.5, however.

Ay.