Re: [gentoo-dev] Doomsayers needed

2014-11-26 Thread Sergey Popov
23.11.2014 21:47, Michael Orlitzky пишет:
 We've got a bug in Nagios's `ping` command format detection:
 
   https://bugs.gentoo.org/show_bug.cgi?id=468296
 
 It's easy to reproduce by taking down your lo interface, or by
 filtering all icmp packets in iptables.
 
 Fortunately, you can override the auto-detection by passing it a magic
 string, and that works around the bug:
 
   --with-ping-command=/bin/ping -n -U -w %d -c %d %s
   --with-ping6-command=/bin/ping6 -n -U -w %d -c %d %s
 
 Those are the formats and executable locations that get detected on my
 machine. Can anyone think of a case where hard-coding these (using
 $ROOT) would backfire on me?
 

Standart - cross-compilation and prefix. If you do not care about the
latter(not having keywords for your package) - it's ok.
Cross-compilation, or compilation into another root is trickier - you
should support it.

-- 
Best regards, Sergey Popov
Gentoo developer
Gentoo Desktop-effects project lead
Gentoo Proxy maintainers project lead



signature.asc
Description: OpenPGP digital signature


Re: [gentoo-dev] Doomsayers needed

2014-11-26 Thread Michael Orlitzky
On 11/26/2014 01:43 PM, Sergey Popov wrote:
 
 Standart - cross-compilation and prefix. If you do not care about the
 latter(not having keywords for your package) - it's ok.
 Cross-compilation, or compilation into another root is trickier - you
 should support it.
 

With ping and ping6 coming from net-misc/iputils, wouldn't the command
syntax be the same under prefix, except with $EROOT instead of $ROOT?

And with the command set to ${ROOT}bin/ping, building for a Gentoo
system under another root should work, right?

Cross-compiling for another system with a different syntax might not
work, of course -- but it didn't before, either. You'd need to use
EXTRA_ECONF for that.




Re: [gentoo-dev] Doomsayers needed

2014-11-26 Thread Diego Elio Pettenò
On 26 Nov 2014 11:07, Michael Orlitzky m...@gentoo.org wrote:

 On 11/26/2014 01:43 PM, Sergey Popov wrote:
 
  Standart - cross-compilation and prefix. If you do not care about the
  latter(not having keywords for your package) - it's ok.
  Cross-compilation, or compilation into another root is trickier - you
  should support it.
 

 With ping and ping6 coming from net-misc/iputils, wouldn't the command
 syntax be the same under prefix, except with $EROOT instead of $ROOT?

 And with the command set to ${ROOT}bin/ping, building for a Gentoo
 system under another root should work, right?

No, $ROOT should not seep into the compiled code.


Re: [gentoo-dev] Doomsayers needed

2014-11-26 Thread Michael Orlitzky
On 11/26/2014 03:57 PM, Diego Elio Pettenò wrote:

 And with the command set to ${ROOT}bin/ping, building for a Gentoo
 system under another root should work, right?
 
 No, $ROOT should not seep into the compiled code.
 

Ah, I think I see my mistake: when running *within* a chroot, you don't
want ${ROOT}bin/ping, instead you want just /bin/ping. Likewise for
prefix I'd want ${EPREFIX}/bin/ping instead of ${EROOT}bin/ping?

As long as the test suite remains disabled, it shouldn't be a problem to
set them to the values needed at runtime. Thank you both.




[gentoo-dev] Doomsayers needed

2014-11-23 Thread Michael Orlitzky
We've got a bug in Nagios's `ping` command format detection:

  https://bugs.gentoo.org/show_bug.cgi?id=468296

It's easy to reproduce by taking down your lo interface, or by
filtering all icmp packets in iptables.

Fortunately, you can override the auto-detection by passing it a magic
string, and that works around the bug:

  --with-ping-command=/bin/ping -n -U -w %d -c %d %s
  --with-ping6-command=/bin/ping6 -n -U -w %d -c %d %s

Those are the formats and executable locations that get detected on my
machine. Can anyone think of a case where hard-coding these (using
$ROOT) would backfire on me?