Bug #74980 for IO-Socket-IP: no error message for non-existent host

Mon Feb 13 19:54:43 2012 zefram [...] fysh.org - Ticket created

Subject:	no error message for non-existent host
Date:	Tue, 14 Feb 2012 00:54:30 +0000
To:	bug-IO-Socket-IP [...] rt.cpan.org
From:	Zefram <zefram [...] fysh.org>

$ perl -MIO::Socket::IP -lwe 'IO::Socket::IP->new(PeerHost=>"doesnt.exist", PeerPort=>123) or die "error: $!"' error: at -e line 1. I was expecting some kind of message in $!. IO::Socket::INET gives "Invalid argument", which isn't great but at least hints at the error. -zefram

Wed Feb 15 05:00:29 2012 leonerd-cpan [...] leonerd.org.uk - Correspondence added

On Mon Feb 13 19:54:43 2012, zefram@fysh.org wrote: Show quoted text

> $ perl -MIO::Socket::IP -lwe 'IO::Socket::IP-

> >new(PeerHost=>"doesnt.exist", PeerPort=>123) or die "error: $!"'

> error: at -e line 1. > > I was expecting some kind of message in $!. IO::Socket::INET gives > "Invalid argument", which isn't great but at least hints at the error.

Since not every error that might stop the constructor can be represented by an errno value, it is the general protocol of IO::Socket and subclasses to use $@ instead to signify an error. This is mostly true of resolver errors; getaddrinfo() does not set $! on failure, it uses its own return code. On the Principle of Least Surprise I suppose the constructor ought to set something by way of an error, though. Perhaps EINVAL is the best choice here, or maybe I can find something else suitably descriptive. -- Paul Evans

Wed Feb 15 05:00:30 2012 The RT System itself - Status changed from 'new' to 'open'

Wed Feb 15 05:00:43 2012 leonerd-cpan [...] leonerd.org.uk - Broken in 0.08 added

Wed Feb 15 05:00:43 2012 leonerd-cpan [...] leonerd.org.uk - Broken in 0.08_004 added

Wed Feb 15 05:12:01 2012 zefram [...] fysh.org - Correspondence added

CC:	zefram [...] fysh.org
Subject:	Re: [rt.cpan.org #74980] no error message for non-existent host
Date:	Wed, 15 Feb 2012 10:11:42 +0000
To:	Paul Evans via RT <bug-IO-Socket-IP [...] rt.cpan.org>
From:	Zefram <zefram [...] fysh.org>

Paul Evans via RT wrote: Show quoted text

>Since not every error that might stop the constructor can be represented >by an errno value, it is the general protocol of IO::Socket and >subclasses to use $@ instead to signify an error.

Oh, hmm. Confusing, since that makes it different from all the other I/O interfaces. Not documented for IO::Socket or IO::Socket::INET. I see IO::Socket::IP does document filling $@, but even it omits to say what it'll actually return on error. Show quoted text

>On the Principle of Least Surprise I suppose the constructor ought to >set something by way of an error, though. Perhaps EINVAL is the best >choice here,

Upon looking at the code, it turns out IO::Socket::INET is deliberately returning EINVAL as well as populating $@. -zefram

Tue May 08 14:19:14 2012 leonerd-cpan [...] leonerd.org.uk - Correspondence added

On Wed Feb 15 05:12:01 2012, zefram@fysh.org wrote: Show quoted text

> Upon looking at the code, it turns out IO::Socket::INET is deliberately > returning EINVAL as well as populating $@.

This behaviour now copied into 0.09: $ perl -Mblib -MIO::Socket::IP -lwe 'IO::Socket::IP->new(PeerHost=>"doesnt.exist", PeerPort=>123) or die "error: $!"' error: Invalid argument at -e line 1. -- Paul Evans

Tue May 08 14:19:15 2012 leonerd-cpan [...] leonerd.org.uk - Status changed from 'open' to 'resolved'

Tue May 08 14:19:23 2012 leonerd-cpan [...] leonerd.org.uk - Fixed in 0.09 added