Bug#674523: apt-get manpage: please document option --solver
On Thu, Aug 18, 2016 at 08:37:00PM +0000, Ximin Luo wrote:
> > I have that patched locally too, that's not a big deal. The big deal is
> > documenting the other options properly and not refer to the .txt file
> > for developers that describes the entire protocol.
> >
>
> What do you mean by "document the other options properly"?
For example,
APT::Solver::Preferences
APT::Solver::Strict-Pinning
should probably be documented in apt.conf if we document --solver.
> > I also have no intention to manipulate the entity files for a simple niche
> > option.
> >
>
> $ sudo apt-get --solver x update
> E: Command line option --solver is not understood in combination with the other options
> 100
>
> From this I thought it was specific to "install", but now I see it's also accepted for some other things like "remove" and "purge". I'll update the patch.
I'm not sure how that is related to what I wrote. What I wrote is very simple:
The options should be documented where they belong, and not taint invocation
examples defined in the .ent files.
> $ sudo apt-get --solver dump install ack-grep
> [..]
> The following packages have unmet dependencies:
> ack-grep : Depends: libfile-next-perl (>= 1.10) but it is not going to be installed
> E: Write error - write (32: Broken pipe)
> E: Write error - write (32: Broken pipe)
> 100
>
> I get also the same result, when running without sudo, directly in a
> root shell.
I read something about a bug somewhere, that it's too fast or something,
David knows more. But it should(TM) work.
>
> By its name "dump", I figured it was an internal component and not an
> actual solver. I guess it is supposed to dump the EDSP to stdout? I
> will update the patch, but correct me if I'm wrong about the previous.
As it's help message says when invoked:
"apt-dump-solver is an interface to store an EDSP scenario in\n"
"a file and optionally forwards it to another solver.
> Don't be so arrogant to believe that your judgment of "absurd" is
> objective or obvious. There was nothing in my original report that
> should have caused a reaction of anger.
It *is* objective *and* obvious. The severities have clearly defined
semantics (which I quoted), and are not defined by your personal
opinions. Feel free to add your own severity usertag.
> Dependency resolution is a significant usability issue, and not
> documenting features that improve this situation, has a major effect
> on other people's perceptions of the whole software. The fix might be
> trivial, but not fixing it has a non-trivial result.
The external solvers do not deliver substantially different results,
unless tweaked manually, which has different syntax per solver,
and might not be documented anywhere.
In most cases, apt dependency solving works just fine. In the usual
cases where it breaks down, like install stuff that depends on non-candidate
versions, the external solvers are not much of a help. By default they
will say no solution. You have to turn of Strict-Pinning first, and then
possibly tweak your solver via Preferences to prefer not to install
stuff from experimental, which is a non-trivial operation for the
average user.
--
Debian Developer - deb.li/jak | jak-linux.org - free software dev
When replying, only quote what is necessary, and write each reply
directly below the part(s) it pertains to (`inline'). Thank you.
Reply to: