Hello everyone,
Forwarding the message to the list as apparently Julien doesn't have
enough time for this currently.
Since then, 0.7.40 has been uploaded to experimental. I may also
provide individual patches for the issues I mentioned below to be
reviewed, if necessary.
Begin forwarded message:
Date: Mon, 4 Mar 2013 21:47:49 +0100
From: Andrew Shadura <bugzilla@tut.by>
To: Julien Cristau <jcristau@debian.org>
Subject: ifupdown in wheezy
Hello,
First of all, let me quote the list of changes I intended to push to
wheezy originally:
[ Andrew Shadura ]
* Don't configure bridge interfaces as tagged VLAN interfaces
(Closes: #696642).
* Add "tryonce" option to DHCP-enabled methods (Closes: #694541).
* Implement inet6/auto for kFreeBSD, call DHCP release of ifdown
on Linux (Closes: #701884).
* Update manual pages.
[ Stéphane Graber ]
* Patches for upstart support from Ubuntu:
- Start the job on runlevel [2345]. This is a no-op during a normal
boot since the network will be started *before* runlevel is
emitted, but is needed to restart the network after a change from
runlevel 1 (LP: #752481).
- Don't bring 'lo' down (add it to --exclude).
- Emit deconfiguring-networking (LP: #1061639).
- Update network-interface-security job to stop when the parent job
is stopped itself. This avoids leftover instances (LP: #1065684).
* Set MTU of tunnel devices (LP: #1074048).
* Actually set the new calculated value for duplicate entries
(LP: #1086517).
[ Josselin Mouette ]
* postinst: Do not create /etc/network/interfaces if it was removed
manually (Closes: #695906)
I understand that probably it's not a good timing to get all of them
now to testing, but some of them are still needed, I think.
Bug #695906 was kind of fixed by NMU uploaded by Michael Gilbert. In my
opinion, that wasn't a good way of solving it, and it wasn't cool he
uploaded without some ack. from me, but okay, that happened.
Of the rest of the changes, I'd like to emphasise few I think we must
have in wheezy.
First of all, VLAN issue, #696642. As you certainly know Debian is very
often used on servers, and it's quite a common setup to have VLANs and
stuff. By not fixing this bug may have lots of people upset, as they
would need to work it around, while the fix is quite easy, tested by
the bug reporter, so I'm sure it works.
LP #1074048 bug is fixable with one-line patch:
- ip link set %iface% up
+ ip link set %iface% up [[mtu %mtu%]]
So I think it's worth including. Or at least I should remove that
option from the manual page so users won't be confused.
LP #1086517 is also easily fixed:
@@ -1342,6 +1342,10 @@
return NULL;
}
+ if ((*var)[j].value == value) {
+ return value;
+ }
+
free((*var)[j].value);
(*var)[j].value = strdup(value);
if (!(*var)[j].value) {
@@ -2144,6 +2148,7 @@
currif->option[i].value[l] = '\n';
strcpy(&(currif->option[i].value[l + 1]), rest);
+ rest = currif->option[i].value;
}
}
}
Same here, if we don't fix it, I should fix the docs.
Speaking of upstart script fixes, we can probably live them out if you
think they're too much at this point.
Bug #701884: fixable with ~6 lines of interface definition code, the
question is if it's important enough for the release, and where should
we fix it if it is: in ifupdown or in d-i? Not sure, but the patch is
ready anyway.
DHCP -1 issue. So what do we do with it? Return behaviour from squeeze,
without -1, leave it there as it's in ~3 Ubuntu releases, add the
option? As patching DHCP client isn't acceptable at this point,
something has to be done.
P.S. All of those changes are about to be uploaded to experimental as
version 0.7.40 so at least they may be tested by users.
--
WBR, Andrew
--
WBR, Andrew
Attachment:
signature.asc
Description: PGP signature