Rob Landley rob at landley.net
Tue Aug 30 20:47:27 UTC 2005

When's the last time our test suite got much attention paid to it?  I just did 
a "make check" and a number of things fail, several of which I'm probably 
responsible for.

One of the ones, which is really odd, is that gnu sort on kbuntu is acting as 
if "-f" is always on.  (It's doing a case insensitive sort on the README.)  I 
don't know why it's doing this.  I'm fairly certain it shouldn't _be_ doing 
this.  And I think under Fedora (where I tested my sort rewrite) it wasn't 
doing this.  But there you have it.  (No, I checked alias, there's no alias 
for sort.  Could be something to do with LANG=en_US.UTF-8 or who knows what 

Anyway, my own tests have generally been written in python, which I haven't 
checked in because the build doesn't use it.  I really don't like the current 
method of having 8 gazillion tiny files because source control systems tend 
to be unhappy about renames and deletes (svn isn't as bad as cvs here, but I 
haven't tried to follow the revision history of a directory that's had a lot 
of files renamed and deleted in it.)

What I'd like to do is make one big test file (per applet) that runs lots of 
tests in an intelligible way, reporting the results for each test.  And 
ideally, I'd like to be able to trade code with the Linux Test Project 
people, because they could use some command line tests and we need to pass 
whatever they've got anyway.

Is there any opinion on this approach, one way or the other?


More information about the busybox mailing list