kde-workspace 4.10.90 ABI breakage in libtaskmanager

[Albert Astals Cid]

El Dilluns, 22 de juliol de 2013, a les 13:03:23, Sebastian Kügler va 
escriure:
> On Sunday, July 21, 2013 17:20:13 Aaron J. Seigo wrote:
> > On Sunday, July 21, 2013 15:50:52 Sebastian Kügler wrote:
> > > We have a release blocker here. If nobody steps up to look into this,
> > > I'll
> > > use git log / git blame and a long pointy stick.
> > 
> > What we have is a change in policy. To recap:
> > 
> > * the libraries in kde-workspace have no ABI or API commitments. none.
> > * the libraries in kde-workspace have changed ABI and API multiple times
> > in
> > the past, and this has not always (ever?) been accompanied by changes of
> > the .so version
> > * modules outside of kde-workspace should NOT be using these libraries,
> > but
> > though several do, in particular libkworkspace [1]
> > 
> > there are a number of things that could be done to improve this situation:
> > 
> > * not installing headers; unfortunately there are users of these libraries
> > and  they will likely complain if we stop installing headers. still, this
> > may be wise to do
> > * bumping .so versions on every single API/ABI incompatible release (which
> > has  been numerous in the 4.x series); given the lack of ABI/API
> > committment, this seems a little academic in nature
> > 
> > personally i’d vote for not installing headers and let builds break.
> > 
> > i’m fine with bumping the .so number of the library as well, though i
> > don’t
> > consider that a compelling solution to the real problem of using libraries
> > outside of kde-workspace that are specifically not intended to be used
> > outside of kde-workspace.
> > 
> > [1] there is one weak exception to this: kdeplasma-addons which is for all
> > intents and purposes a kde-workspaces sub-module
> 
> I suppose we had installed the header already in previous releases? In that
> case, it's probably wiser (given that we're at RC stage) to just bump the
> .so version and revisit this situation for Plasma2?


Agreed, let's either:
 a) Fix it by bumping the soname
 b) Ignore it

Don't break other people's code at RC2 stage.

Cheers,
  Albert