Re: [Evolution-hackers] e-d-s ABI breakage [ .so bump ] ...
- From: Harish Krishnaswamy <kharish novell com>
- To: michael meeks novell com
- Cc: Srinivasa Ragavan <SRagavan novell com>, evolution <evolution-hackers gnome org>, Caolan McNamara <caolanm redhat com>, Robert Love <rml ximian com>, Veerapuram Varadhan <VVaradhan novell com>
- Subject: Re: [Evolution-hackers] e-d-s ABI breakage [ .so bump ] ...
- Date: Wed, 16 Aug 2006 14:12:14 +0530
On Mon, 2006-08-14 at 16:12 +0100, Michael Meeks wrote:
> So,
>
> I spent a while digging into this and trying to patch it to remove the
> changes and create something that could be back-compat, so we could
> downgrade the .so version again.
>
> Then - I noticed that (apparently) SL 10.1 and SLED10 shipped with the
> new ABI anyway; but (presuambly) we just failed to bump the .so number;
> at least - I'm giving up until Wed. on the grounds of
> incomprehensibility - but it looks like:
>
The ebook ABI change has been committed to the 2.7.x series and not for
2.6.x which ships on SLED10 / SL 10.1.
> * abi breakage occured *
> * SL10.1/code-10 ships *
> * .so versions bumped
> - now -
> * Gnome 2.16 ships *
>
> Which makes it look like code-10 shipped with a libebook that (while
> having a different .so number) is fully compatible with the current ABI,
> (and yet incompatible with the previous version with the same .so
> number).
[Addressbook]
AFAICT, none of the ABI break changes in ebook were pushed into
autobuild and the version on SLED10 is not ABI compatible with the one
shipping with 2.7.90 and onwards. I will poke Varadhan [on CC] who
packages Evolution for SLED10 and get back to you.
> I guess that turns the problem into (mostly) a SUSE issue that we can
> work around by some duplication/linking of the the various libraries
> twice in our packages - [ugh]; and of course - reverting the ABI
> breakage wouldn't help us - it'd prolly just further confuse an already
> messy situation.
>
> Unless I'm confused again ?
[Calendar] Clock Applet (Panel) already uses the changed APIs for
handling recurrence data [1] and no other application in SLED10 is
affected by this - the libecal SONAME has not changed and it is binary
compatible with Evolution 2.7.91. Technically - this is akin to zero
breakage on SONAMES, rpm deps et all and code-wise, the impact has
already been handled.
[1] JP had put in the fix to the panel.
-Harish
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]