-----BEGIN PGP SIGNED MESSAGE-----
On 04/24/2013 06:11 PM, Simo Sorce wrote:
> On Wed, 2013-04-24 at 20:55 +0200, Sumit Bose wrote:
>> On Wed, Apr 24, 2013 at 02:22:42PM -0400, Stephen Gallagher
>>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>>> On Wed 24 Apr 2013 02:05:55 PM EDT, Michal Židek wrote:
>>>> New version of the patch is attached.
>>>> Changes from previous iteration: - version-info set to 0:2:0
>>>> - init function remains backward compatible, but it sets
>>>> default oadptions needed for range calculation. Non default
>>>> values can be set via ded setters. - sss_idmap_opts moved to
>>>> private header
>>> This is still a backwards-incompatible break. If we were
>>> presenting a data structure in a public header and no longer
>>> are, it's an ABI break. Is there any way to avoid this?
>> sss_idmap_opts is introduced with this patch, it was in the
>> public header in the first version and Michal moved it into the
>> private one in the new version.
> I think Steven was referring to struct sdap_idmap_slice
> disappearing from src/providers/ldap/sdap_idmap.h maybe ?
Actually, I hadn't looked at the patch. I saw that he had moved a
structure from a public to a private header and didn't realize it was
a newly-added structure.
One nack: the version-info string needs to be 1:0:1 according to
This is because you have added the getters and setters to the public
interface, which means that all of steps 3, 4 and 5 in that guideline
must be followed, which results in a value of 1:0:1 for version-info.
Note: this arrangement does NOT result in a SOname bump, because the
"age" is equal to the "current" (Which effectively means that
saying that this interface is a proper superset of all interfaces that
have preceded it).
Thanks a lot for the explanation Stephen. I was actually following the
same guide, but I thought that only one rule is applied each time
version-info is changed, which did not make much sense in some
situations. Now it finally makes sense to me.
New patch is attached.