dhcpcd-discuss

On 2016-01-31 04:04, Nick Williams wrote:
> > I'll pick that out :)
> > What is an important bug?
>
> I guess what I was asking here is, “is there a version before which
> DHCPv6 and/or prefix delegation just don’t work at all because of
> some major bug that was fixed?” I get that we need to perform
> thorough testing, etc.

OK, here's a little bug for you :)
There appears to be an issue with assigning delegated prefixes to 
interfaces with a prefix length of 64 or greater right now.
This issue appears to have always been there and I'm working on a fix in 
my spare time.
As no-one has noticed it recently (although it could also be Peng's 
issue) I assume it's not urgent.
However, it c

There is also an issue where the config is not working for multiple 
interfaces to delegate to.
This issue appears to have been recently introduced, but I can't say 
from which version.

> But, since we’re looking at having to package
> dhcpcd ourselves for our version of Debian, it would save me time if I
> knew that there’s no point in trying anything before version X.

This is why I provide a list of important changes with each release to 
this mailing list.
You can read the archives here:
http://roy.marples.name/archives/dhcpcd-discuss/

> That aside, I have a couple more questions:
>
> 1) What’s the point of config options “ipv4” / “ipv6” in
> comparison to / conjunction with “dhcp” / “dhcp6” (slash their
> “noipv4,” etc. companions)? Why would you use just “ipv4” or
> just “dhcp” or both “ipv4” and “dhcp”? I read this [1]
> documentation as well as I could, but that just stuck out to me as not
> making sense.

IPv4 also always static IP assignment, ARP ing profiles and IPv4LL 
support - not just dhcp.
IPv6 also has RS/RA - not just dhcpv6.

> 2) What will “allowinterfaces" do if you pass to it a non-existent
> interface? For example, if I set the option “allowinterfaces plc0”
> (short for placeholder 0), will it, as I hope, keep dhcpcd from
> touching any existing interfaces, or will it error out because plc0
> doesn’t exist?

Here's the text:
 allowinterfaces pattern
    When discovering interfaces, the interface name must match pattern 
which is a space or comma separated list of patterns passed to 
fnmatch(3). If the same interface is matched in denyinterfaces then it 
is still denied.

Now, I think this is quite clear, but let me break it down into smaller 
chunks:
1 Interfaces are discovered by probing the kernel.
2 If allowinterfaces is set, each interface discovered must match the 
pattern passed to fnmatch(3)
3 Because fnmatch(3) states you are allowed to use wildcards it follows 
that an actual pattern doesn't exit but what it can match will exist
4 Therefore, if allowinterfaces has a pattern that does not exist, it's 
not an error.

> What about if I set the option “allowinterfaces”
> without anything after it?

You can an error.
If pattern was optional it would display as [ pattern ], following the 
same format as every other man page out there.

Roy