=head1 METHODS
-=head2 C<< new command => $command, args => \@args >>
+=head2 C<new>
+
+ my $tvc = Test::Valgrind::Command->new(
+ command => $command,
+ args => \@args,
+ );
Creates a new command object of type C<$command> by requiring and redispatching the method call to the module named C<$command> if it contains C<'::'> or to C<Test::Valgrind::Command::$command> otherwise.
The class represented by C<$command> must inherit this class.
sub new_trainer { }
-=head2 C<args $session>
+=head2 C<args>
+
+ my @args = $tvc->args($session);
Returns the list of command-specific arguments that are to be passed to C<valgrind>.
sub args { @{$_[0]->{args} || []} }
-=head2 C<env $session>
+=head2 C<env>
+
+ my $env = $tvc->env($session);
This event is called in scalar context before the command is ran, and the returned value goes out of scope when the analysis ends.
It's useful for e.g. setting up C<%ENV> for the child process by returning an L<Env::Sanctify> object, hence the name.
sub env { }
-=head2 C<suppressions_tag $session>
+=head2 C<suppressions_tag>
+
+ my $tag = $tvc->suppressions_tag($session);
Returns a identifier that will be used to pick up the right suppressions for running the command, or C<undef> to indicate that no special suppressions are needed.
sub suppressions_tag;
-=head2 C<filter $session, $report>
+=head2 C<filter>
+
+ my $filtered_report = $tvc->filter($session, $report);
The C<$session> calls this method after receiving a report from the tool and before forwarding it to the action.
You can either return a mangled C<$report> (which does not need to be a clone of the original) or C<undef> if you want the action to ignore it completely.