There is a canonpath case where we reduce redundant directories down to
nothing but should instead reduce them to the magical six zeros root
directory. There have been two different definitions of "nothing" in
recent history.
Before File::Spec 0.89:
$ perl -"MFile::Spec" -e "print File::Spec->canonpath('d0:[d1.-]foo.dat');"
d0:[]foo.dat
Starting with 0.89:
$ perl -"MFile::Spec" -e "print File::Spec->canonpath('d0:[d1.-]foo.dat');"
d0:foo.dat
These are both wrong, though the MakeMaker test infrastructure preferred
the old wrong way to the new wrong way. What we should see (and do after
the attached patch) is:
$ perl -"MFile::Spec" -e "print File::Spec->canonpath('d0:[d1.-]foo.dat');"
d0:[000000]foo.dat
which compares favorably to what the native filename syntax parser does:
$ write sys$output f$parse("d0:[d1.-]foo.dat")
D0:[000000]FOO.DAT;
--- lib/File/Spec/VMS.pm;-0 Mon Nov 1 10:03:24 2004
+++ lib/File/Spec/VMS.pm Thu Nov 11 09:15:52 2004
@@ -182,7 +182,7 @@ sub canonpath {
$path =~ s/\.[^\]\.]+\.-\./\./g; # .foo.-. ==> .
$path =~ s/\[[^\]\.]+\.-\./\[/g; # [foo.-. ==> [
$path =~ s/\.[^\]\.]+\.-\]/\]/g; # .foo.-] ==> ]
- $path =~ s/\[[^\]\.]+\.-\]/\[\]/g; # [foo.-] ==> []
+ $path =~ s/\[[^\]\.]+\.-\]/\[000000\]/g;# [foo.-] ==> [000000]
$path =~ s/\[\]//; # [] ==>
return $path;
}
--- lib/File/Spec/t/Spec.t;-0 Sun Sep 5 17:29:24 2004
+++ lib/File/Spec/t/Spec.t Thu Nov 11 08:52:52 2004
@@ -322,6 +322,7 @@
[ "VMS->canonpath('[d1.000000.][000000.][d3.--.000000]file.txt')",
'[d1.000000]file.txt' ],
[ "VMS->canonpath('volume:[d1.000000.][000000.][-.-.000000]file.txt')",
'volume:[000000]file.txt' ],
[ "VMS->canonpath('[d1.000000.][000000.][--.-.000000]file.txt')",
'[-.000000]file.txt' ],
+[ "VMS->canonpath('[d1.d2.--]file')", '[000000]file'
],
[ "VMS->splitdir('')", '' ],
[ "VMS->splitdir('[]')", '' ],
