[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#841347: packages are not marked as auto "i A" in aptitude



Control: reassign -1 aptitude 0.8.3-1

On Wed, Oct 19, 2016 at 07:04:03PM +0200, Michael Biebl wrote:
> I typically use mk-build-depends -i to install the build dependencies of
> a package.
> 
> Those $foo-build-depends pile up and from time to time I use aptitude
> (in interactive mode) to remove them. But none of the dependencies are
> actually removed then. Afaics, they are all in state "i" instead of
> "i A".
> 
> Interestingly, if I remove the $foo-build-depends package via apt, the
> packages are properly marked for autoremoval.
> 
> Is there some broken interaction between aptitude and apt regarding the
> auto state? I vaguely remember that this worked in the past.
> 
> If you want to reproduce the problem, try the following, with the
> attached deb:
> 
> $ apt file install ./gtk+3.0-build-deps_3.22.1-4_amd64.deb

(the "file" in these commands is bogus)

> [install lots of stuff]
> $ apt remove gtk+3.0-build-deps
> [lists lots of packages as autoinstalled]
> $ apt autoremove
> [removes lots of stuff]
> 
> $ apt file install ./gtk+3.0-build-deps_3.22.1-4_amd64.deb
> [install lots of stuff]
> $ aptitude remove gtk+3.0-build-deps
> [removes gtk+3.0-build-deps]
> $ apt autoremove
> 0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.


As far as I can see in the reproduce steps apt tools are working as
expected, it is aptitude which isn't working to your liking, hence
reassigning as I have no (real) idea about aptitude.

Given that apt autoremoves the dependencies I guess they are correctly
marked as automatic on installation (you can check with 'apt-mark
showauto' to be sure), but somehow aptitude looses the autobit on
removal. aptitude would usually by default remove the autoremoveable
packages together with the -build-deps package, wouldn't it?

I /guess/ that could be intended behavior depending on how you are using
aptitude to remove the package. The introtext mentions interactive mode,
your reproducer would be command mode. Especially in the first one every
keypress can count, so perhaps you want to expand on what you do – but
aptitude people will know better than me what to do (of course, in the
event that this is a libapt regression feel free to hand it back).


Best regards

David Kalnischkies

Attachment: signature.asc
Description: PGP signature


Reply to: