Re: pbulk resolving stage ALWAYS failes

[Aleksey Cheusov <cheusov%tut.by@localhost>]

 >> No. I have lots of local patches but neither of them concern dr

> The message tells you that archivers/sarab can't depend on
> archivers/dar. As I don't see this on a different machine, this means
> either (a) Your tree has some chagnes in archivers/sarab we don't see or
> (b) Your tree has some changes in archivers/dar that we don't see or
> (c) The scan hasn't seen archivers/dar.

> You could start by grepping pscan for "dar-", you should get at least a
> PKGNAME line and the dependency line of sarab.

In order to continue bulk build I already edit pscan by hand.
So, I'll see it again at the time of the next bulk build.

 >> I repeat: failure of 1 of 9500 packages leads to complete fall.
 >> What a...?

> I give the person responsible a change to fix it. Otherwise I will
> request a backout the offending changes. The point here is that an
> unresolvable dependency is a major issue and MUST NOT be ignored.
Every type of bugs MUST NOT be ignored. And all types of building
failures have the same same priority just because the result is the
same: package does not build. That's all.

> We had such issues in the past and they are ignored for month
> because noone cared as noone saw them.
Bulk build framework MUST provide a visible answer on question:
why this package fails.

> This is an even bigger issue
> on the non-primary platforms.
The are TONS of really idiotic bugs that are not seen on not Net- and
DragonFlyBSDs. 

One of my favourite:
http://mova.org/~cheusov/pub/pkgsrc-pbulk/Debian-etch/current/log/20071210.1550/gzip-1.3.12

So what? All they are the same including bad or missed dependancy,
all of them should be seen by developers and all of them MUST NOT be ignored.
But neither of them should stop bulk build.

-- 
Best regards, Aleksey Cheusov.