Philip M. Gollucci wrote:
Fred Moyer wrote:
http://people.apache.org/~phred/Apache-Bootstrap-0.02.tar.gz
As I tested this out with Apache::Dispatch today, I came to the
realization that Apache::Bootstrap itself must be bootstrapped in the
Makefile.PL. Here's how I did it in Apache::Dispatch, parts ripped
off from ExtUtil::MakeMaker. I'm still fuzzy on the details of how
not to cause CPAN testers failures.
Any thoughts here welcome - still bootstrapping this (no pun intended :)
BEGIN {
my $AB_VER = 0.02;
my $fail_msg = "Warning: prerequisite Apache::Bootstrap $AB_VER
not found.";
eval { require Apache::Bootstrap };
die $fail_msg . "\n" if $@;
die $fail_msg . " We have $Apache::Bootstrap::VERSION.\n"
if $Apache::Bootstrap::VERSION < $AB_VER;
}
... some code
my $mp_gen = Apache::Bootstrap->satisfy_mp_generation(2);
I haven't read the code more than you put here, but this doesn't seem
quite right.
It seems cyclic. I know perl is one of those languages that let you get
away with it if you are clever, but I don't know many cases where its
right idea....
Bootstrapping is cyclic to some degree, but I don't think this code is
quite where I want it to be. I have my head fairly well wrapped around
most of the issues now. This module is really intended to provide
information about existing mod_perl and apache::test installs so that
the build program (Makefile.PL) can make decisions on whether to use
ExtUtils::MakeMaker (mp1) vs ModPerl::MM (mp2).
I think for simplicity the bootstrapping stuff should remain thin. I
looked at some of the CPAN tester failures for my modules, and I am
following the path of making build failures due to dependencies result
in NA cpan tester report, vs fail. If the dependencies aren't
installed, the test should not be run the result should be NA as opposed
to Fail. If you want to see details of what I mean, go find the test
results for RHP::Timer on cpan testers and you'll see NA entries where
the required version of perl is not installed, as opposed to Fail.
I'm going to make another iteration and have the bootstrap process fail
at compile time if it cannot find any mod_perl installation, just like
build will fail if the required version of perl cannot be found.
Without a mod_perl install, we should assume the module cannot be built.
Does that sound like a good assumption?
I haven't done any of the work though, so I'll defer to you unless I get
my act in gear.
Thanks for the feedback, it is helpful. If this works, it should make
shipping dual life modules fairly painless (crosses fingers).
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
