File::Spec - catfile wrongly makes relative path absolute #8760
Comments
From [email protected]Created by [email protected]Consider: If the first element in the list passed to File::Spec->catfile The behaviour is similar on other architectures as well, in that On Windows, "\blib" is returned. Solaris returns "/ blib" (with IMHO, leading and trailing empty (or undef) elements should be File::Spec->catdir has the same bug. Solution (e.g. for File::Spec::Unix.pm): ---snip--- Inline Patch--- Unix.pm 2006-10-11 19:11:03.000000000 +0200
+++ Unix.pm.new 2007-02-01 21:51:22.000000000 +0100
@@ -78,6 +78,7 @@
sub catdir {
my $self = shift;
+ shift @_ while @_ && $_[0] =~ /^$/;
$self->canonpath(join('/', @_, '')); # '' because need a trailing '/'
}
@@ -91,6 +92,7 @@
sub catfile {
my $self = shift;
+ shift @_ while @_ && $_[0] =~ /^$/;
my $file = $self->canonpath(pop @_);
return $file unless @_;
my $dir = $self->catdir(@_);
---snip---
--shmemPerl Info |
From @schwernGeorg Moritz (via RT) wrote:
I do not believe this is a bug but an underdocumented feature of cat*. The intention is to mirror the behavior of splitdir() which uses an empty string to indicate the root directory. This allows splitdir and catdir/file to round trip. Additionally, "" really is the true name of the root directory. Consider it this way and it will make more sense. catdir() takes a list of directory names. "/" is a path separator and not part of the directory name. With that in mind, how do you represent the root directory? What is its name? We call it / but really its name is the empty string. The directory-which-shall-not-be-named. The directory formerly known as "/". This is not just a File::Spec thing but an accurate way of looking at the structure of the Unix filesystem. Consider it another way. Given that the point of File::Spec is to be filesystem agnostic, how would you put together an absolute path? You can't refer to it as "/", that's Unix specific. That it accepts undef this is simply because undef becomes an empty string when used as a string. Turn on warnings. |
|
The RT System itself - Status changed from 'new' to 'open' |
From [email protected]Created by [email protected]Consider: If the first element in the list passed to File::Spec->catfile The behaviour is similar on other architectures as well, in that On Windows, "\blib" is returned. Solaris returns "/ blib" (with IMHO, leading and trailing empty (or undef) elements should be File::Spec->catdir has the same bug. Solution (e.g. for File::Spec::Unix.pm): ---snip--- Inline Patch--- Unix.pm 2006-10-11 19:11:03.000000000 +0200
+++ Unix.pm.new 2007-02-01 21:51:22.000000000 +0100
@@ -78,6 +78,7 @@
sub catdir {
my $self = shift;
+ shift @_ while @_ && $_[0] =~ /^$/;
$self->canonpath(join('/', @_, '')); # '' because need a trailing '/'
}
@@ -91,6 +92,7 @@
sub catfile {
my $self = shift;
+ shift @_ while @_ && $_[0] =~ /^$/;
my $file = $self->canonpath(pop @_);
return $file unless @_;
my $dir = $self->catdir(@_);
---snip---
--shmemPerl Info |
|
@rgs - Status changed from 'open' to 'rejected' |
From @adamkennedyCan I strongly recommend that before you make any changes here you add Because as I understand it, "" intentionally represents the root path. DB<4> x File::Spec->splitdir('/foo/bar') DB<5> x File::Spec->catdir(File::Spec->splitdir('/foo/bar')) I seem to recall I rely on this "null string is root" in a number of As for the undef case, I'm ambivalent, but I'd rather see an error than Adam K Georg Moritz wrote:
|
Migrated from rt.perl.org#41422 (status was 'rejected')
Searchable as RT41422$
The text was updated successfully, but these errors were encountered: