Re: SONAME bumping and d-i
On Thu, Dec 16, 2004 at 02:44:43PM -0500, Andres Salomon wrote:
> This is from #debian-boot, regarding bumping the SONAME and reverting the
> CAN-2004-081 patch (see #284356 for details):
> <dilinger> joeyh: have you talked to horms about his incrementing of the
> 2.4.27 SONAME stuff yet?
> <joeyh> I haven't. I've seen his mail
> <joeyh> if ya gotta do it, ya gotta do it, but it will of course cause a
> certian amount of d-i pain
> <dilinger> he's trying to decide whether he should revert the patch in -1,
> and make a -2
> <dilinger> well, at this point...
> <dilinger> a) we've already broken the ABI. to revert it will break the
> ABI again, and
> <dilinger> b) if it's a large hassle for d-i to switch kernel package names,
> is it really worth reverting? i assume d-i will want to use the kernel w/
> security patches
> <dilinger> we'll probably have to do the same for 2.6, of course
> <joeyh> oh, so you're talking about us just continuing in with the broken abi
> <dilinger> yes, if we think that's the better solution. at this point, i
> don't know
> <joeyh> it seems to me that either approach will probably end up breaking rc2
> at some point.
> <joeyh> not changing the soname seems more likely to me to break it though
> <dilinger> ok
> <joeyh> if the package name changes, we have to update rootskel and the
> initrds, iirc that should be all
> <dilinger> so you're a fan of bumping the SONAME and reverting for -1, then
> <joeyh> hmm, let me think about it some more
> <joeyh> if we bump the soname, the kernel udebs package names will also change
> <dilinger> ok. i'm going to quote this and post to the (d-k) list, just
> follow up w/ what you decide. horms and i talked about it on irc last night,
> it would be good to get a discussion on the list
Andres, thanks for passing this my way.
Joey, thanks for the indirect feedback,
it is very much appreciated.
My position is that the patch probably should be included.
However it is still pending incusion upstream and from that
I deduce that 1) it is a bit scary and 2) the world doesn't fall
appart if you don't have it. Please see, http://lkml.org/lkml/2004/12/15/299
With that in mind, while I would still like to see the patch
included and thus bumping to -2, I do accept that it may well be
more pain than it is worth. As far as I see there seem to three main
options being discussed at this time, Joey and the d-i team, I am
really looking to you to help decide which way to go.
1. Release a new set of -1 binary packages with the ABI changed reversed
This means an updated kernel-source-2.4.27 and an updated
2. Release a set of -2 binary packages with the ABI change included
Again, this means an updated kernel-source-2.4.27 and an updated
kernel-image-2.4.27-<arch>. I believe that it also means that
the -1 binary packages will end up being removed from the archive.
3. Do 1, then 2.
The thing I am not sure about here, is that if releasing -2 removes
the -1 packages from the archive, is there any point to doing this?
Perhaps there is a delay, say for some d-i milestone to be reached
between 1 & 2, in that case it might make some sense.
A few additional points.
* There seems to be a second ABI change relating to this patch.
This isn't a big deal, as long as it gets rolled into the ABI change
if we choose option 2 or 3.
* None of the other changes that were included in kernel-source-2.4.27
2.4.27-6 or are pending for 2.4.27-7 break the ABI as far as I can see.
And I did look pretty hard.