On Sun, 3 Sep 2006 02:04:08 +0200
bardo <[EMAIL PROTECTED]> wrote:

> 2006/9/3, Jason Chu <[EMAIL PROTECTED]>:
> > On Sat, 2 Sep 2006 18:17:01 +0200
> > How would your file be different from anyone else's?  A developer's
> > perhaps?
> 
> It wouldn't. But I have 533 packages in my cache, and it took almost
> three hours to check them all. It would not be the first time that the
> devs benefit from something like this without having to download all
> the packages and do all the computational work by themselves...
> 
> > You know you could just run namcap /var/cache/pacman/pkg/* >
> > ~/badpackages.log...
> 
> Uh, didn't know. I got used to pacman's syntax, which not always does
> accept multiple names... thanks ;-)
> 
> > As long as developers are doing their jobs, they should have run
> > namcap on the package before pushing it to the server.  Shouldn't
> > that information be the same?
> 
> Yeah, they should've done it, but it doesn't seem everyone has done
> so. I've got a really huge file (10958 lines, 3685 if we remove the
> ones related to non-standard directories [still 34 packages] and to
> testing packages), and some entries just shouldn't be ignored. A few
> stats:
> -138 included but already satisfied dependencies
> -497 included and not needed deps (many false positives here for
> unrecognized deps, mostly for interpreted languages)
> -471 libtool occurrencies... and I doubt they're all necessary
> -186 deps detected and not included (false positives here sometimes)
> 
> Do I need to go on? I'm not blaming the developers, I think it is
> something that happens normally with time. It just needs to be fixed,
> and the devs are the ones who can do it. If the help of the community
> is needed, then we could call a dep-fixing day, rather then a
> bug-squashing one...

I think specific problems reported to the bug tracker would help the
developers.  A wholesale scanning of all packages won't really help
cause none of the developers will want to sift through it and figure
out what problems are problems and what aren't...

> > I've always wanted to keep namcap a simple rule-based package
> > checker. I wouldn't want to sully its simplicity with an exception
> > db.  Also, who would maintain it?  How often would it be updated?
> > Who says what messages are non-needed?
> 
> Yeah, that was just an idea. I don't really like it, too, but
> discussion is good, maybe we'll come up with some good ideas.
> You're right, it'd be really difficult to maintain, but I don't see
> any point in advising the dev that /home/httpd is not a standard dir:
> he knows it, he's the one who did it, and he did on purpose. This is
> surely not enough, but there's more: it *is* standard for apache under
> arch, isn't it? So why bother?
> 
> Another idea I don't like but may inspire someone to find something
> with more sense: every package could have a namcap-targeted file that
> says: yeah, I know you think I need gcc, but believe me, I don't, and
> I know that those files seem to be where they shouldn't, but leave
> them alone.
> The problem of this approach is the lack of extensibility of arch's
> package format, which is somewhat strict with the metadata, I think to
> understand...

It's true, we don't have any room for metadata like that right now.
Though, it could be added.  My fear is that they would be abused and
the assumptions wouldn't be checked ie. if a package *did* depend on
gcc that next version, but the developer told it to ignore that
message, how would they even know to check?

> Anyway, thanks for your answer =)

Jason

-- 
Two years ago, I generated a gpg key that expired August 24, 2006
(686FD13C).  I've now generated a new one (C7EFE9F4).

Attachment: signature.asc
Description: PGP signature

_______________________________________________
arch mailing list
[email protected]
http://www.archlinux.org/mailman/listinfo/arch

Reply via email to