Le 6 avr. 10 à 16:16, Eugene V. Lyubimkin a écrit :
Thibaut Paumard wrote:So, the _real_ relationship is: the conjunction (B(old) + C(new)) breaksConflicts should be declared only in case of file conflicts. If, as I see,A and A-doc (to a point where A-doc ends-up deconfigured).Very clearly, C (new) must declare a "Conflicts" or a "Breaks" against B(< new). But which one is the good choice?it's the case of 'doesn't work with' then use Breaks.
This does make sense :-)I'm a little confused because here it's not only B which is broken (but a larger subsystem). If I read "Breaks: B" not as "breaks B" but as "breaks things in conjunction with B", then I guess I get the point.
I'm a little worried by this part from policy:"dpkg will refuse to allow the package which declares Breaks be installed unless the broken package is deconfigured first".
In my case, due to the triggers mechanisms, bad things will happen _during the dpkg run_ if dpkg launches the trigger when the files of B and C are present (even if B is deconfigured). Is this something that is possible, or will dpkg make sure B is upgraded before triggering update-A-doc? I get the impression the right thing to do would be to deconfigure _A-doc_ until B has been upgraded...
Regards, Thibaut.