[Tcllib-devel] Tcllib testsuite work in the recent past ...

[Andreas K. <aku...@sh...>]

Hi ya all.

People might have seen the series of large commits I did in the last
days and week. This mail is just a heads up of what I did, and why.

After a long time of procrastinating on this I finally managed to get
the motivation to dive into the Tcllib testsuite as a whole and give
it a nice scrubbing.

For example, while most of the testsuites for the various modules and
packages had the correct code to load the local sources for testing
instead of from (external) installation we did still had a number of
packages which did it wrong, or only half-right. And those which did
it completely right not only had heaps of ugliness coming before the
actual set of tests, but the code in question was in essence
replicated all over.

Half of the work I did was to draw the common parts out of this code
and put them into a separate file to be shared by all testsuites. This
additionally had the benefit of hiding all the ugly machinery behind
commands with (nearly) self-explanatory names.

The other half then was just drudgery, replacing all the existing
replicated boilerplate with much small prologues using all the shiny
new stuff.

As an end result the headers of all the .test files should now be much
more readable when it comes to declaring the needs of the contained
testsuite, be it dependencies on versions of Tcl and tcltest, or the
loading of the package under test and whatever other supporting files
and packages are needed.

It should make the creation of new testsuites for Tcllib easier as
well.


Now, while I consider the main cleanup of the testsuites to be done
there are still places in there which could be better. Namely:

- A lot of the testsuites under "struct" declare helper commands, some
  of them quite large. We should consider to put them into their own
  file and then declare that as a support.

- The testsuites under "math" have similar problems. They are a bit
  worse as well IMHO, as their helpers are also duplicated across the
  testsuites. And there seem to be several different implementations
  of some of them floating around.

  One helper for example compares two floating point numbers for
  equality within 4 digits after the decimal point. I know I have
  seen two different implementations of that, and I believe that there
  might actually be three.

- A number of packages have different implementations as well, pure
  Tcl vs. one or more accelerators. Only some of them have logic to
  allow the switching between implementations at runtime. And very few
  of the testsuites actually test all of the implementations they can
  can lay their hands on as well.

  There is opportunity for refactoring and generalization, for both
  package implementations, and accompanying testsuites.


So, even with this cleanup now behind us there are still lots of
things we can do to make Tcllib better and more polished. Lets go
forward.

-- 
So long,
	Andreas Kupries <aku...@sh...>
			<http://www.purl.org/NET/akupries/>
	Developer @	<http://www.activestate.com/>
-------------------------------------------------------------------------------