[pacman-dev] [PATCH] Reduce the negligence of _alpm_resolvedeps

[Xavier]

On Sat, Dec 01, 2007 at 12:24:16AM +0100, Nagy Gabor wrote:
> > On Thu, Nov 29, 2007 at 10:13:42PM +0100, Nagy Gabor wrote:
> > > OK, but I leave the benchmark for you;-)
> > > Note: Long "dependency paths" like in smoke001.py may cause notable
> > > slowdown. I made a quick "hack", so probably I made some mistakes...
> > > 
> > 
> > I couldn't measure any difference. pacman always need ~3.5 seconds for
> > running smoke001.
> > 
> Because it is an add transaction;-) I just said that it contains a long
> dependency path, which can cause notable slowdown in case of -S
> first_element_of_path. (Unfortunately I cannot write such tricky
> pactests in python).
> 

Oops, my bad. In my defense, it was a bit late :)
But now, I'm slightly confused. How hard is it to edit smoke001 to do what
you want? It looks like a totally trivial change to me. So I'm not sure I
understood what you wanted.

> > > From 3b8f21423035f7bf683494a8e369b6501be8e379 Mon Sep 17 00:00:00
> > > 2001 From: Nagy Gabor <ngaba at bibl.u-szeged.hu>
> > > Date: Thu, 29 Nov 2007 21:55:55 +0100
> > > Subject: [PATCH] Reduce the negligence of _alpm_resolvedeps
> > > 
> > > This is a not-to-commit patch
> > > NOTE: -Se is temporarily removed, because now resolvedeps doesn't
> > > have spkg param
> > > 
> > 
> > What is really bad about this patch besides breaking -Se?
> > Do you see a clean way of reimplementing it?
> Well, the way you didn't like earlier:
> http://projects.archlinux.org/git/?p=pacman.git;a=commit;h=b0c064d59b8786a1e169e305a263fc13e2d7ed32
> 
> dependency-list == populate(populatedlist - oldlist)
> 

The previous code did that, and you wanted to call recursively sync_prepare
on dependency-list, right?
I think that's why I didn't like it much.

In my opinion, there is no need to worry much about -Se. Where is this option
used anyway?
If there are use cases for it, it probably only involes running -Se with one
target. Not several targets with inter-dependencies. So this extreme corner
case doesn't need to be supported.
That is, sync1001 should still pass, but sync1002 doesn't need to.