user32:input.c fails only when run from make?!

[Dan Kegel]

On Wed, Sep 24, 2008 at 12:50 PM, Dan Kegel <dank at kegel.com> wrote:
>... On my Levnovo x60 running Hardy,
> the user32 test input.c fails when run from make
> nearly all of the time, but doesn't fail if run standalone.

FWIW, running the test under valgrind makes it fail reliably with

input.c:715: Test failed: a5/2: the msg 0x0101 was expected, but got msg 0x0105

on several machines.   It'd be nice if the test was robust enough to pass
in that (slow) environment.