Bug #47802 for Module-Manifest: -f prevents file /dev/null

Sat Jul 11 21:26:02 2009 user42 [...] zip.com.au - Ticket created

Subject:	-f prevents file /dev/null
Date:	Sun, 12 Jul 2009 11:25:24 +1000
To:	bug-Module-Manifest [...] rt.cpan.org
From:	Kevin Ryde <user42 [...] zip.com.au>

With Module::Manifest 0.07 I tried use Module::Manifest; my $manifest = Module::Manifest->new('/dev/null', '/dev/null'); and got an error Did not provide a readable file path at foo.pl line 2 whereas of course /dev/null does exist and is readable. In $manifest->open I think the -f test should be omitted. I believe it almost never helps to pre-empt what open() might be capable of on a given file/device/socket/etc (at least once using three-arg open or whatever to ensure the name is not interpreted ...)

Sun Jul 12 20:48:01 2009 adam [...] ali.as - Correspondence added

Subject:	Re: [rt.cpan.org #47802] -f prevents file /dev/null
Date:	Mon, 13 Jul 2009 10:47:42 +1000
To:	bug-Module-Manifest [...] rt.cpan.org
From:	Adam Kennedy <adamkennedybackup [...] gmail.com>

Non-files are host-dependant and can't be reliably packaged into distributions in any platform-neutral manner. I'm also curious why you'd want to package a null device into a tarball... Adam K 2009/7/12 Kevin Ryde via RT <bug-Module-Manifest@rt.cpan.org>: Show quoted text

> Sat Jul 11 21:26:02 2009: Request 47802 was acted upon. > Transaction: Ticket created by user42@zip.com.au > Queue: Module-Manifest > Subject: -f prevents file /dev/null > Broken in: (no value) > Severity: (no value) > Owner: Nobody > Requestors: user42@zip.com.au > Status: new > Ticket <URL: https://rt.cpan.org/Ticket/Display.html?id=47802 > > > > With Module::Manifest 0.07 I tried > > use Module::Manifest; > my $manifest = Module::Manifest->new('/dev/null', '/dev/null'); > > and got an error > > Did not provide a readable file path at foo.pl line 2 > > whereas of course /dev/null does exist and is readable. > > In $manifest->open I think the -f test should be omitted. I believe it > almost never helps to pre-empt what open() might be capable of on a > given file/device/socket/etc (at least once using three-arg open or > whatever to ensure the name is not interpreted ...) > >

Sun Jul 12 20:48:02 2009 The RT System itself - Status changed from 'new' to 'open'

Mon Jul 13 21:09:16 2009 user42 [...] zip.com.au - Correspondence added

Subject:	Re: [rt.cpan.org #47802] -f prevents file /dev/null
Date:	Tue, 14 Jul 2009 11:08:39 +1000
To:	bug-Module-Manifest [...] rt.cpan.org
From:	Kevin Ryde <user42 [...] zip.com.au>

"Reserved Local Account via RT" <bug-Module-Manifest@rt.cpan.org> writes: Show quoted text

> > I'm also curious why you'd want to package a null device into a > tarball...

No, I was trying it outside, with an empty manifest file, and what better empty file than /dev/null :-). (The convenience of the -f operator has a lot to answer for. Sometimes -e is reasonable, usually both just make perl code more restricted in what it will operate on than similar C code where the tedium of stat() dissuades anyone checking before opening. :-)

Tue Jul 14 06:13:22 2009 adam [...] ali.as - Correspondence added

Subject:	Re: [rt.cpan.org #47802] -f prevents file /dev/null
Date:	Tue, 14 Jul 2009 20:13:03 +1000
To:	bug-Module-Manifest [...] rt.cpan.org
From:	Adam Kennedy <adamkennedybackup [...] gmail.com>

In this case, the behaviour was actually as intended. Will resolve closed. Adam K 2009/7/14 Kevin Ryde via RT <bug-Module-Manifest@rt.cpan.org>: Show quoted text

> Queue: Module-Manifest > Ticket <URL: https://rt.cpan.org/Ticket/Display.html?id=47802 > > > "Reserved Local Account via RT" <bug-Module-Manifest@rt.cpan.org> writes:

>> >> I'm also curious why you'd want to package a null device into a >> tarball...

> > No, I was trying it outside, with an empty manifest file, and what > better empty file than /dev/null :-). > > > (The convenience of the -f operator has a lot to answer for. Sometimes > -e is reasonable, usually both just make perl code more restricted in > what it will operate on than similar C code where the tedium of stat() > dissuades anyone checking before opening. :-) > >

Wed Jul 15 19:37:52 2009 user42 [...] zip.com.au - Correspondence added

Subject:	Re: [rt.cpan.org #47802] -f prevents file /dev/null
Date:	Thu, 16 Jul 2009 09:37:18 +1000
To:	bug-Module-Manifest [...] rt.cpan.org
From:	Kevin Ryde <user42 [...] zip.com.au>

"Reserved Local Account via RT" <bug-Module-Manifest@rt.cpan.org> writes: Show quoted text

> > as intended.

?? It's one of those things where less code makes more features. Eg. someone making a shell pipeline can use /dev/stdin when a front-end insists on a filename. But if you do want to avoid char-specials or the like then of course testing before opening is a race condition, the test should be on the opened handle.