[Bug 2076655] [NEW] [FFe] APT 3.0 solver updates

[Julian Andres Klode]

Public bug reported:

I'm asking for a blanket FFe for the experimental 3.0 solver that can be
used with the --solver 3.0 argument. It's mostly really bug fixing
that's happening, I'm not sure what would constitute a feature at this
point*, but the changes are still fairly big as I'm refactoring it to
align more with the MiniSAT paper; that fixes bugs along the process.

* If you consider we don't have conflict-driven clause learning which
makes it slow, but is it a feature? Right now it doesn't terminate in my
real life mantic -> noble upgrade test case, this seems an essential bug
fix.

This FFe is restricted to the solver3 code in apt-pkg/solver3.{cc,h} and
does not cover modifications to other APT parts.

In any case given this is an undocumented experimental option we don't
have to be worried about regression potential.

** Affects: apt (Ubuntu)
     Importance: Undecided
         Status: New

** Description changed:

  I'm asking for a blanket FFe for the experimental 3.0 solver that can be
  used with the --solver 3.0 argument. It's mostly really bug fixing
  that's happening, I'm not sure what would constitute a feature at this
  point*, but the changes are still fairly big as I'm aligning it with the
  MiniSAT paper more.
  
  * If you consider we don't have conflict-driven clause learning which
  makes it slow, but is it a feature? Right now it doesn't terminate in my
  real life mantic -> noble upgrade test case, this seems an essential bug
  fix.
+ 
+ One thing we don't yet have which is really a feature which is out of
+ scope for 24.10 is the logic for installing replacement packages (where
+ it allows replacing foo with bar even if foo is manually installed, but
+ bar Conflicts/Replaces/Provides it).

** Description changed:

  I'm asking for a blanket FFe for the experimental 3.0 solver that can be
  used with the --solver 3.0 argument. It's mostly really bug fixing
  that's happening, I'm not sure what would constitute a feature at this
- point*, but the changes are still fairly big as I'm aligning it with the
- MiniSAT paper more.
+ point*, but the changes are still fairly big as I'm refactoring it to
+ align more with the MiniSAT paper; that fixes bugs along the process.
  
  * If you consider we don't have conflict-driven clause learning which
  makes it slow, but is it a feature? Right now it doesn't terminate in my
  real life mantic -> noble upgrade test case, this seems an essential bug
  fix.
- 
- One thing we don't yet have which is really a feature which is out of
- scope for 24.10 is the logic for installing replacement packages (where
- it allows replacing foo with bar even if foo is manually installed, but
- bar Conflicts/Replaces/Provides it).

** Description changed:

  I'm asking for a blanket FFe for the experimental 3.0 solver that can be
  used with the --solver 3.0 argument. It's mostly really bug fixing
  that's happening, I'm not sure what would constitute a feature at this
  point*, but the changes are still fairly big as I'm refactoring it to
  align more with the MiniSAT paper; that fixes bugs along the process.
  
  * If you consider we don't have conflict-driven clause learning which
  makes it slow, but is it a feature? Right now it doesn't terminate in my
  real life mantic -> noble upgrade test case, this seems an essential bug
  fix.
+ 
+ In any case given this is an undocumented experimental option we don't
+ have to be worried about regression potential.

** Description changed:

  I'm asking for a blanket FFe for the experimental 3.0 solver that can be
  used with the --solver 3.0 argument. It's mostly really bug fixing
  that's happening, I'm not sure what would constitute a feature at this
  point*, but the changes are still fairly big as I'm refactoring it to
  align more with the MiniSAT paper; that fixes bugs along the process.
  
  * If you consider we don't have conflict-driven clause learning which
  makes it slow, but is it a feature? Right now it doesn't terminate in my
  real life mantic -> noble upgrade test case, this seems an essential bug
  fix.
  
+ This FFe is restricted to the solver3 code in apt-pkg/solver3.{cc,h} and
+ does not cover modifications to other APT parts.
+ 
  In any case given this is an undocumented experimental option we don't
  have to be worried about regression potential.

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to apt in Ubuntu.
https://bugs.launchpad.net/bugs/2076655

Title:
  [FFe] APT 3.0 solver updates

Status in apt package in Ubuntu:
  New

Bug description:
  I'm asking for a blanket FFe for the experimental 3.0 solver that can
  be used with the --solver 3.0 argument. It's mostly really bug fixing
  that's happening, I'm not sure what would constitute a feature at this
  point*, but the changes are still fairly big as I'm refactoring it to
  align more with the MiniSAT paper; that fixes bugs along the process.

  * If you consider we don't have conflict-driven clause learning which
  makes it slow, but is it a feature? Right now it doesn't terminate in
  my real life mantic -> noble upgrade test case, this seems an
  essential bug fix.

  This FFe is restricted to the solver3 code in apt-pkg/solver3.{cc,h}
  and does not cover modifications to other APT parts.

  In any case given this is an undocumented experimental option we don't
  have to be worried about regression potential.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/apt/+bug/2076655/+subscriptions