Bug#883619: Any reason not to simply upload ceres-solver with adjusted version of libeigen3-dev

[Andreas Tille]

Hi Philipp,

On Tue, Feb 26, 2019 at 10:45:52AM +0100, Philipp Huebner wrote:
> 
> > I mean: There is no *effective* change since the Build-Depends we set is
> > fullfilled anyway by the existing packages.  Its just that it is
> > explicitly declared in the package metadata.
> 
> IMO that's a valid fix for buster, but does not fix the underlying issue
> for sid / buster+1.

Sure, but we want to release Buster and if this bug is not fixed
ceres-solver (and its dependencies are removed in now 4.3 days.  Thus
I'm just building the current HEAD and will upload if nobody will stop
me.
 
> The next time a newer version of eigen3 is uploaded, you'll have the
> same "problem" all over again: adjust the build dep and upload to get a
> corresponding rebuild.
> Updating eigen3 renders ceres and similar packages unusable until
> rebuilt, please read through the discussion in #868355 to understand
> what I mean.
> 
> The eigen3 maintainer and I are happy to simply rebuild affected
> packages after every eigen3 update, but Emilio considers it an upstream bug.
> Unfortunately I could not find anybody able to shed more light on the
> eigen3 topic.

I agree that the topic seems to be more complex in general but for the
moment we need a fix for Buster and that fix is very simple - so I do
not see any reason to not fix it.  You might like to reopen the relevant
bugs (I mean #868355 - I just asked for closing which was done and
#883619) with lower severity to keep on discussing for Buster+1.

Kind regards

       Andreas.

-- 
http://fam-tille.de