[Bash-completion-devel] Bump match_max for test suite?

[Crestez Dan Leonard]

On Sat, 2010-01-09 at 13:02 +0100, Guillaume Rousse wrote: 
> I guess it's because you have a large rpm db, right ? Should we not try
> to provide a constant test environement, instead of relying of tester
> one ? It may be difficult in some case (such as creating a fake rpm db
> here), but others (such as host completion relying on content of
> HOSTFILE) should be more easy to mock.

Isn't that unreasonably hard? There are completions for a lot of
programs; some of which can even conflict (for example dpkg and rpm). It
makes sense to run as many tests as possible and report what was not
tested. Assuming multiple people run the tests on different systems
regressions should still be caught.

A test could simply check that 'loc<TAB>' complete into
'localhost' (among possibly others). I could then report untested if the
system has no entry for localhost. For some programs such assumptions
would make it a easier to write more complicated tests. I believe it's
acceptable to include various assumptions in tests as long as it doesn't
generate spurious failures. Is that OK?