Bug#687694: version incompatibility

[Niels Thykier]

On 2012-10-04 22:45, brian.thomason at gmail.com wrote:
> Hi,
> 

Hi,

> I'm happy to make whatever changes are necessary to this package, but I'm
> not sure what is gained by providing Breaks for every potential package
> affected by a bouncycastle upgrade.  It seems to me more proper that if
> package X depends on bouncycastle = version Y (which appears to be the case
> here) that package should then reflect it, rather than putting X number of
> Breaks in the bouncycastle package itself and having to expand that list
> each time a new package enters the pool that depends upon it.
> 
> Best regards,
> 
> -Brian
> 
> [...]

It would be optimal if reverse dependencies would have the proper upper
bound, but fact is that they don't.  And at this point it is too late to
introduce the upper bound (and relying on it).
  The use of breaks in bouncycastle is a (temporary) fail-safe, in case
someone does a partial upgrade including bouncycastle (from sid), but
without (e.g.) libitext-java.  The libitext-java in stable has no upper
bound, so this would be permitted and "break" libitext-java.

To be honest, I am not sure to what extend the breaks are needed, but
libitext-java in Wheezy still does not have the proper version (waiting
for the RT to get back to me).  So partial upgrades between sid and
Wheezy are /still/ broken without a Breaks in bouncycastle on libitext-java.

If we had a decent transition system for Java (e.g. like the one we have
for C/C++ libraries) and it had been followed, we would not have had to
resolve to breaks either.

~Niels