At 03:42 PM 2/19/2001 -0500, [EMAIL PROTECTED] wrote:
>On Mon, Feb 19, 2001 at 08:20:38AM -0600, Jarkko Hietaniemi wrote:
>> Maybe there should be a common source from
>> which t/TEST (or the moral equivalent) and vms/test.com are generated?
>
>Looking at vms/test.com... eek. This is going to be interesting.
What scared you most, the echo replacement written in assembler or the
diverging Perl source based on t/TEST circa 1994 ;-)? I don't know of any
reason to change the OS-specific wrapper, but the Perl part (beginning with
DECK/DOLLAR) should probably be scrapped and the current version of t/TEST
ported to VMS, then run from test.com rather included wholesale.
It remains to be seen whether we can get the current t/TEST to work on VMS
with an acceptable number of "if $^O eq 'VMS'" conditionals or whether we'll
have to continue to maintain (or fail to maintain) a separate file. On the
one hand, Perl on VMS is a lot smarter than it was in 1994 so the port may
not be as much of an uphill battle as it was then, but on the other hand
there are still some VMS-specific gotchas that will need to be accounted
for. Among the wacky things handled by what's in test.com is the case where
the "not ok" message is split across two lines, i.e., one of the things that
turned up broken for Peter Prymmer when he was running tests outside of the
test.com context.
Does anyone on vmsperl have some time to take a look at this?
