[darcs-devel] [darcs #216] memory and time testsuites

[David Roundy via RT]

Fri Feb 11 07:43:09 2005: New ticket: 216.
Transaction: Ticket created by droundy
       Queue: Darcs
     Subject: memory and time testsuites
       Owner: Nobody
  Requestors: droundy at abridgegame.org
      Status: new
 Ticket <URL: http://bugs.darcs.net/.//Ticket/Display.html?id=216 >

It would be great to have test suites (or a single test suite) that would
measure darcs' performance both in terms of peak memory usage and time
spent.  Could be written in perl or something like that, and I don't think
it need be portable--I doubt one can measure memory usage in a portable
fashion.  This would be extremely useful for darcs optimizers:
optimization can be tedious enough without having to write a script to
create large repositories.

I imagine writing a set of perl routines to generate repositories, and then
to generate changes to such a repository, so one could generate a large
repository with a long history with just a little loop of code.  I'd like
to have (as long as I'm dreaming) separate tests for as many darcs commands
as possible, and for the most interesting ones (get, pull, record,
whatsnew) I'd like to have enough different tests to excersize different
parts of the code.  e.g. for record:

 * No changes in a repository with many (small, or ordinary-sized) files.
 * No changes with --look-for-adds in a repository with many files.
 * An initial record in a repository with many files.
 * An initial record in a repository with many binary files.
 * A record with many separate changes (as in many hunk patches) to a
   single file.
 * A record with a few changes to each of many files.
 * A record after darcs replace has been called on many files.

That's all that comes to mind for record.

Ideally, the test suite would have a portable pseudorandom number generator
so the generated repositories would be identical on different platforms and
computers.

David