Re: [gentoo-project] Re: Call for agenda items - Council meeting 2014-08-12

[Rich Freeman]

On Thu, Jul 31, 2014 at 3:12 PM, Michał Górny <mgorny@g.o> wrote:
>
> Yes, exactly. We need to get dynamic-deps right if they are ever
> supposed to become the default. That's one of the reasons that we want
> to revert the problematic changes and make Portage use the default
> model once again.

Do we actually have some kind of list of issues with dynamic deps?
The only specific one that I think I've seen is with prerm and subslot
deps, but as was pointed out that issue actually is as much of a
problem with static deps unless you unmerge all the reverse-deps
before upgrading anything, followed by a re-merge.

I have no problem with accepting that it is broken, but it would be
nice to deal with more specifics and not with it in general.

>
> If you are really curious, I am working hard on providing tools to fix
> the vdb inconsistencies caused by dynamic-deps. There were no specific
> data because it wasn't available until today.
>
> My regularly updated desktop system (2-3 days between @world updates)
> after disabling dynamic-deps has 77 packages needing rebuild. That
> number includes a few virtuals, Perl packages and other low-effort
> cases. And this is after the big, scary virtual/*udev changes.
>
> Over the next days I will obviously have more numbers. More
> specifically, the number of packages needing rebuild after dependency
> changes made by developers. It should be noted that the above number
> includes one-time rebuild of packages that are simply ancient.
>
> There is a lot of FUD about unnecessary rebuilds. Sadly, most people
> seem to fight a holy war against them without realizing the real
> impact. In fact, more unnecessary rebuilds are caused by unnecessary
> USE flags than by dependency changes. Yet the same people believe in
> adding more flags to contain even more minor aspects of packages...

Thank you for this.  It is very helpful in gauging the likely impact
of having more revbumps.

One thing I don't want to do is create a barrier to anybody who wants
to upgrade an eclass or do work on virtuals.  I can just imagine
endless debates about whether splitting a virtual is worth it since it
will cause up to 250 rebuilds, etc.

Is there any easy way to compare tree vs installed deps using the API?
 I have a script that is part of the way there, but I'm struggling to
compare the vartree and porttree depdendencies (the portree
dependencies need to be correctly reduced - if somebody has the list
of function calls needed to reduce an RDEPEND from porttree to account
for USE/etc in /etc/portage and substitute in subslots that would be
helpful.

code snippet:
    depstr = vartree.dbapi.aux_get(cpv, ["RDEPEND"])[0]
    depstr2 = porttree.dbapi.aux_get(cpv, ["RDEPEND"] )[0]
    flatdeps=portage.dep.flatten(portage.dep.paren_reduce(depstr))
    flatdeps2=portage.dep.flatten(portage.dep.use_reduce(portage.dep.paren_reduce(depstr2)))

Ideally those should be the same if static deps = dynamic deps, but I
suspect it isn't appying the right USE flags (I'm not passing any),
and it isn't substituting actual subslot values either.  I'm also not
sure if flatten is going to properly handle operators/etc.

Rich