Re: Issues with library version numbers

[Elijah Newren <newren gmail com>]

On 8/26/05, Mark McLoughlin <markmc redhat com> wrote:
> Hi,
>         Sorry - that was all a bit over the top.
> 
>   1) I think you should add a caveat to your bugs: "maybe these weren't
>      every public functions, that needs to be checked out"
> 
>   2) I think we need to be even more clear about what we mean by ABI
>      compatibility. We don't include accidently exported symbols, we
>      don't include non-public APIs, we only maintain compatibility
>      between stable releases etc.

Yes.  Although Brian's email problem caused you some undue stress (I
definitely was really worried when I suddenly saw lots of libraries
with claimed ABI breakage too), it appears that this will lead to some
better understanding which will hopefully have the side-effect of (a)
helping him and Bala make more correct abicheck scripts (Thanks for
taking that on Brian!) and (b) helping us know where we need to
improve documentation.

>   3) Since GNOME maintainers work hard to ensure ABI compatibility is
>      broken, you should work on the assumption that ABI compatibility
>      hasn't been broken unless you've got *much* more concrete evidence
>      that it has been broken.
> 
>   4) We need to figure out a GNOME-wide policy on whether its
>      worthwhile to bump the mico shared library version and, if we do,
>      when do we do it? Each time a tarball is released with new
>      functions or just once a GNOME release cycle?

I'd love to know the answer to that--and not just for stable
libraries.  Reasons:  I honestly don't know what to do with libwnck
despite the basic instructions in configure.in (yeah, it's not stable,
but I'm trying to figure out when to bump the soname at the right
times to indicate added API as well as when to indicate API breaks)...
 I've had a complaint already on what I did do, from Rodrigo I think. 
And if I don't know how to do this with the modules I work on, I don't
know how to help other modules or verify whether something is wrong if
someone does bring up a problem to r-t or d-d-l.