Bug #19932 for Sys-Hostname-Long: Should not be calling `su` (fails under fakeroot)

Fri Jun 16 06:01:07 2006 JMEHNLE [...] cpan.org - Ticket created

Subject:

Should not be calling `su` (fails under fakeroot)

When $< (UID) is 0, Sys::Hostname::Long does `su nobody -c "hostname -- fqdn"`. Under fakeroot however (e.g. when building and testing packages), this makes `su` ask for a password because the process isn't _really_ running as root. Sys::Hostname::Long shouldn't be calling `su` in the first place!

Fri Jun 16 23:46:04 2006 scottp [...] dd.com.au - Correspondence added

Subject:	Re: [rt.cpan.org #19932] Should not be calling `su` (fails under fakeroot)
Date:	Sat, 17 Jun 2006 03:40:06 +0000 (UTC)
To:	bug-Sys-Hostname-Long [...] rt.cpan.org
From:	Scott Penrose <scottp [...] linux.dd.com.au>

----- Julian Mehnle via RT <bug-Sys-Hostname-Long@rt.cpan.org> wrote: Show quoted text

> Fri Jun 16 06:01:07 2006: Request 19932 was acted upon. > Transaction: Ticket created by JMEHNLE > Queue: Sys-Hostname-Long > Subject: Should not be calling `su` (fails under fakeroot) > Broken in: 1.3, 1.4 > Severity: Important > Owner: Nobody > Requestors: JMEHNLE@cpan.org > Status: new > Ticket <URL: http://rt.cpan.org/Ticket/Display.html?id=19932 > > > > When $< (UID) is 0, Sys::Hostname::Long does `su nobody -c "hostname > -- > fqdn"`. Under fakeroot however (e.g. when building and testing > packages), this makes `su` ask for a password because the process > isn't > _really_ running as root.

Ok I will review that. Show quoted text

> Sys::Hostname::Long shouldn't be calling `su` in the first place!

In an ideal world yes :-) But there are some operating systems out there wich just won't return unless we call "hostname" and if we don't change the user first, then in some operating systems it changes the hostname to "--fqdn". I do believe we could improve it, do more host type specific testing, and rely less on falling back to a sys call (which I really don't like either). Ta Scott

Fri Jun 16 23:46:05 2006 The RT System itself - Status changed from 'new' to 'open'

Sat Jun 17 18:31:16 2006 JMEHNLE [...] cpan.org - Correspondence added

From:

julian [...] mehnle.net

Show quoted text

> > Sys::Hostname::Long shouldn't be calling `su` in the first place!

> > In an ideal world yes :-) But there are some operating systems out > there wich just won't return unless we call "hostname" and if we > don't change the user first, then in some operating systems it > changes the hostname to "--fqdn". > > I do believe we could improve it, do more host type specific testing, > and rely less on falling back to a sys call (which I really don't > like either).

I think there are plenty other options. Take a look at what Sys::Hostname::FQDN and Net::Domain are doing. I know that portably getting the full hostname can be a pain in the ass, but there /really/ should be a better solution than `su nobody -c "hostname --fqdn"`.

Sat Jun 17 21:13:50 2006 scottp [...] dd.com.au - Correspondence added

Subject:	Re: [rt.cpan.org #19932] Should not be calling `su` (fails under fakeroot)
Date:	Sun, 18 Jun 2006 01:03:41 +0000 (UTC)
To:	bug-Sys-Hostname-Long [...] rt.cpan.org
From:	Scott Penrose <scottp [...] linux.dd.com.au>

----- Julian Mehnle via RT <bug-Sys-Hostname-Long@rt.cpan.org> wrote: Show quoted text

> Queue: Sys-Hostname-Long > Ticket <URL: http://rt.cpan.org/Ticket/Display.html?id=19932 > >

> > > Sys::Hostname::Long shouldn't be calling `su` in the first place!

> > > > In an ideal world yes :-) But there are some operating systems out > > there wich just won't return unless we call "hostname" and if we > > don't change the user first, then in some operating systems it > > changes the hostname to "--fqdn". > > > > I do believe we could improve it, do more host type specific

> testing,

> > and rely less on falling back to a sys call (which I really don't > > like either).

> > I think there are plenty other options. Take a look at what > Sys::Hostname::FQDN and Net::Domain are doing. > > I know that portably getting the full hostname can be a pain in the > ass, > but there /really/ should be a better solution than `su nobody -c > "hostname --fqdn"`.

Agreed, except... Note that the above is the last resort. Sys::Hostname::Long goes through heaps of ways of doing it, same as the Sys::Hostname does. Although, I wrote Sys::Hostname::Long quite a long time ago, and FQDN was written many years after that, so maybe if they are doing it better it should be retired or combined. Scott