cpanm detects and does not try to double-install modules that depend on themselves #503
Conversation
tomeon
force-pushed
the
prereq_cycle
branch
3 times, most recently
from
b67ecf6
to
09738f3
Compare
Menlo-Legacy/lib/Menlo/CLI/Compat.pm
Outdated
| my $module_name = $self->find_module_name($configure_state) || $dist->{meta}{name}; | ||
| $module_name =~ s/-/::/g; | ||
|
|
||
| my @all_deps = $self->find_prereqs($dist); | ||
| my @deps = grep { $_->module ne $module_name } @all_deps; |
There was a problem hiding this comment.
Will depending on SelfDep::SubPackage from the same dist trick this check?
There was a problem hiding this comment.
Indeed it would; well-spotted. I was concerned about this as well, but thought it better to catch at least some cases rather than none.
I've pushed some changes to address the sub-package case: now, the code uses Module::Metadata->package_versions_from_directory to enumerate all packages in $dist->{dir}. This should catch the SelfDep::SubPackage case, but I'm worried that the logic is now overbroad and catches things like modules that aren't exposed for public consumption (e.g. test-suite-only stuff).
There was a problem hiding this comment.
I'm worried that the logic is now overbroad and catches things like modules that aren't exposed for public consumption (e.g. test-suite-only stuff).
On reflection, this probably isn't a big deal.
There was a problem hiding this comment.
Yep. You can hide from the metacpan indexer with:
package # hide from indexer
MyApp::Secret;I assume the same trick works elsewhere
|
Some tests are failing, but they're also failing in |
Some modules list themselves as dependencies (looking at you,
Bundle::CPAN). In combination withcpanm's tracking of$self->{seen}{$module}, this causescpanmto mark such a module as having an unsatisfied dependency -- namely, itself. This PR implements a fix that filters out a module from the list of its dependencies and issues a warning message to that effect.