Ken Irving wrote:
> I saw a note on clpm mentioning Expect v1.08, so upgraded to it, but
> I find that my scripts using v1.07 fail to work properly. I'm in the
> process of tracking down the difficulty(/ies), and could let you know
> what I find, if that would be helpful.
Yes, that would be nice, so that I either can fix the incompatibilities
or at least document them. Could you send me some short example
script(s)
that's failing now? Maybe I can spot obvious problems. I'm also trying
to build up a basic testsuite, and backward-compatibility is
important.
> The Changes file says very little, only a mention of "multi-matching",
> which I take to be multiline matching; the docs indicates that this
> defaults to 1, so I tried changing it to 0, thinking this would revert
> to v1.07 behavior, but my problems remain.
No, the "multi-match" functionality is something else: you can
now expect on several spawned programs at once. There is also
a new callback interface that allows to write quite complex
state machines in a single expect command. Take this example
how to automate login:
$obj->expect($timeout,
[
qr'login: $',
sub {
my $fh = shift;
print $fh "$username\r"; exp_continue;
}
],
[
'Password: $',
sub {
my $fh = shift;
print $fh "$password\r"; exp_continue;
}
],
[
timeout =>
sub {
die "No login.\n";
}
],
'-re', qr'[#>:] $', #' wait for shell prompt, then exit expect
);
This is much more elegant than the old interface where you
had to do a switch after the expect to see which pattern matched.
> Perhaps it would be better to increment the version number by a bit more
> than .01 to reflect the extent of changes made?
In hindsight, this would have been better.
> FWIW, I'm using Linux 2.2.5 system, and still using perl 5.005_02 (maybe
> that's the problem?).
No, I developed the new version with 5.00502, so that's OK.
It's simply that the whole matching engine got twisted
a lot, and this breaks backward compatibility.
Hope this helps,
Roland
--
RGi...@cp...
|
