Re: [Bug-dejagnu] dejagnu-1.4 "make check"

[Rob Savoye]

On Thu, Apr 05, 2001 at 01:28:24PM -0400, Ken Raeburn wrote:

> After doing "configure" and "make", but before installing, I tried
> "make check".  It printed warnings because it couldn't find "runtest",
> and finished up with a zero exit status.  This should, IMHO, be
> considered a fatal error.  And if it's meant to be testing the new

  This should be a fatal error. I haven't seen this behaviour on any
of the machines I tested 1.4 on, but I'll try to reproduce it. "make check"
is supposed to be a quite sanity test of DejaGnu before installation. I
rewrote most of the DejaGnu testsuite for the 1.4 release, but it looks
like I need to tweak it alittle more...

> clear to me if it was looking for runtest to drive a test suite, in
> which case you'd want an installed version, or to test the one it
> found, in which case, it shouldn't need to be installed.)
 
  Usually the RUNTEST support in Automake, and many other GNU programs
tests to see whether if should use a local runtest, or the installed one.

> I installed and put the bin directory into $PATH, and tried "make
> check" again.  It got a nonzero exit status because the calc test
> reported a failure.  Which, near as I can tell, was the intended
> effect because calc has an intentional bug.  So, as a test of calc, it
> should be a failure, but as a test of dejagnu, it should be a success.
 
  Correct. I'll fix this. Thanks for the feedback. Other than this, does
it run tests (like lets say the GCC tests) on NetBSD ok ?

        - rob -