Bug #47141 for namespace-clean: Other entries in the glob are not restored properly

Fri Jun 19 04:50:18 2009 ben [...] morrow.me.uk - Ticket created

Subject:	Other entries in the glob are not restored properly
Date:	Fri, 19 Jun 2009 09:49:37 +0100
To:	bug-namespace-clean [...] rt.cpan.org
From:	Ben Morrow <ben [...] morrow.me.uk>

namespace::clean currently doesn't restore the other entries in the glob properly when it deletes a function. The attached test program demonstrates the failure. Fixing the obvious error: --- clean.orig 2009-06-19 09:41:16.000000000 +0100 +++ lib/namespace/clean.pm 2009-06-19 09:41:31.000000000 +0100 @@ -166,9 +166,11 @@ next SYMBOL if $store->{exclude}{ $f }; no strict 'refs'; + local *__tmp; + # keep original value to restore non-code slots { no warnings 'uninitialized'; # fix possible unimports - local *__tmp = *{ ${ "${cleanee}::" }{ $f } }; + *__tmp = *{ ${ "${cleanee}::" }{ $f } }; delete ${ "${cleanee}::" }{ $f }; } makes some but not all of the failures go away; most annoyingly, at least here the SCALAR slot always seems to end up getting wiped. I don't really know what's going on here: it's quite possible the remaining failures are actually bugs in perl. Ben

Message body is not shown because sender requested not to inline it.

Sat Jun 20 10:08:26 2009 ben [...] morrow.me.uk - Correspondence added

Subject:	Re: [rt.cpan.org #47141] Other entries in the glob are not restored properly
Date:	Sat, 20 Jun 2009 15:08:05 +0100
To:	Bugs in namespace-clean via RT <bug-namespace-clean [...] rt.cpan.org>
From:	Ben Morrow <ben [...] morrow.me.uk>

OK, I've worked out what's going on: *x = *y{FORMAT}; is putting the FORMAT ref into the SCALAR slot by mistake. Since formats aren't much used, I think the best thing to do is to just forget about coying the format across, and avoid clobbering the scalar. The attached patch does this, documents that fact, and includes a modified version of the previous tests. Ben

Message body is not shown because sender requested not to inline it.

Tue Jun 30 01:12:17 2009 rafl [...] debian.org - Correspondence added

That modified testcase passes, even without your namespace::clean modifications. Is that intentional?

Tue Jun 30 01:12:19 2009 The RT System itself - Status changed from 'new' to 'open'

Tue Jun 30 07:24:55 2009 ben [...] morrow.me.uk - Correspondence added

Subject:	Re: [rt.cpan.org #47141] Other entries in the glob are not restored properly
Date:	Tue, 30 Jun 2009 12:24:16 +0100
To:	Florian Ragwitz via RT <bug-namespace-clean [...] rt.cpan.org>
From:	Ben Morrow <ben [...] morrow.me.uk>

At 1AM -0400 on 30/06/09 you (Florian Ragwitz via RT) wrote: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=47141 > > > That modified testcase passes, even without your namespace::clean > modifications. Is that intentional?

...you're right, so it does. No, it wasn't intentional. t/lib/OtherTypes.pm should have sub foo { 1 } use namespace::clean; rather than the other way around. Sorry about that. Investigating this perl bug a little further, I've found that it's only present on 5.8.9 and 5.10.0: 5.10.1 will have a fix when it comes out. It might be worth using something like use constant CAN_COPY_FORMAT => do { local *foo, *bar; eval "format foo=\n.\n"; my $sc = *bar{SCALAR}; *bar = *foo{FORMAT}; *bar{SCALAR} == $sc; }; to test if FORMAT can be copied or not. Ben

Fri Jul 29 03:21:22 2011 ribasushi [...] leporine.io - Fixed in 0.12 added

Fri Jul 29 03:21:22 2011 ribasushi [...] leporine.io - Correspondence added

Has been fixed a while ago (0.12)

Fri Jul 29 03:21:25 2011 ribasushi [...] leporine.io - Status changed from 'open' to 'resolved'