Re: cvs commit: src/etc rc rc.shutdown rc.subr src/etc/rc.d localpkg src/sys/sys param.h

From: Doug Barton <dougb_at_FreeBSD.org>
Date: Fri, 16 Dec 2005 13:43:58 -0800
Jeremie Le Hen wrote:

> When I first read your mail on arch_at_ I didn't think about it, but now
> I'm reading this thread, it stirs my mind.  RELENG_6 is supposed to
> be very stable.  Though the ultimate stability would be to stay in
> RELENG_6_0, I think MFC'ing those changes to RELENG_6 would cause much
> pain to system administrators who will be upgrading from 6.0 to 6.1 if
> they don't not update their apache13 port accordingly.

Thank you for putting more thought into this topic. :)

Your concerns are valid, and that is why the code was changed shortly after 
I introduced it into HEAD to run all scripts in a subshell, which is what 
was happening prior to the changes being introduced. The only changes that 
are present now are to the order in which the scripts are run at boot time, 
and while there have been no reports of problems since the change I 
mentioned above was made, any problems that do exist in that area will need 
to be addressed in any case, since that is fundamental to what is being 
added (the ability to run local boot scripts in the base rcorder).

There is another aspect to this that you should be aware of. A group of 
stalwart developers have stepped forward and agreed to work on converting 
the boot scripts that are still in the old style, and helping to fix any of 
the new style scripts that have problems. They are also working on changes 
to the ports infrastructure to handle installing the boot scripts as foo 
instead of foo.sh on RELENG_6 and HEAD systems that have already had the 
update. Meanwhile, while this work is in progress, all scripts will still be 
run in a subshell, so there is virtually zero danger that a wonky script 
will cause the boot to fail.

I hope that this information helps you feel more comfortable with this 
change. I agree that we want to keep FreeBSD's standards high, and that we 
don't want to violate the release engineering principles that our users have 
come to rely on. That is why this change will not be ported back to 
RELENG_5. On the other hand, this is still early enough in the release cycle 
for RELENG_6, and there is enough time between now and when the 6.1 release 
will be cut to ferret out any remaining problems, that we feel it's safe to 
move forward with this. And, if it turns out that I'm wrong, we can easily 
back the changes out in RELENG_6 and keep this change on hold for the 
7.x-RELEASE cycle.

Regards,

Doug


-- 

     This .signature sanitized for your protection
Received on Fri Dec 16 2005 - 20:44:03 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:49 UTC