Prelink success story :)
Toshio
toshio at tiki-lounge.com
Thu Feb 26 23:25:03 UTC 2004
On Thu, 2004-02-26 at 15:09, Dag Wieers wrote:
> Then you clearly have much more time than I have.
>
Yep! I'm unemployed so there's no sense trying to out-talk me; only
out-reasoning will do :-)
>
> > > 2] I have many packages that _have_ to change the %setup line,
> > > 230 of the 622 spec-files which is over 30% (remember perl-packages ?)
> > Doesn't matter. I took a look at several of your perl spec's.
> > They do:
> > %setup -n %{rname}-%{version}
> > which will get caught by #2 above.
>
> And you said you hadn't seen any ocassions where %setup -n was needed
What I said was:
%setup -n foo-oldver [I've never seen this construct, only
%{name}-%{version} which will fail b/c #2] or the tarball doesn't
include versions in its toplevel directory [I have seen this])
I can see that you might have misinterpretted it. Mea Culpa.
If I had written "I've never seen 'foo' and 'oldver' hardcoded into
a %setup macro, only %setup -n %{name}-%{version} which will fail"
it would have been clearer.
> I understand you wanted to know the number of
> packages that have '-n' used and not %{version}. Still 87 do, about 13%.
>
Excellent. Good data. So over 1 in 10 packages can potentially get
past the packager and have to be caught by the QA people. Not trivial.
> Although I must say I don't see why that would be of any value in the
> discussion.
>
Because no other package will make it past the rpmbuild stage with
mismatched version and Source0. So only these are potential QA
problems.
> > > 3] I don't rely on QA people as I'd rather automate and assume a
> > > QA person has better things to do.
> > That's fine. But your question was whether the QA person would catch
> > the problem...
>
> Well, we will not know, would we. I'm just stating it's useless to ask
> this from a QA person if you can automate it.
True.
Coupled with Ville's comment that package QA people really
should be checking out web pages and so forth to make sure they have
a canonical source rather than cut 'n paste, macros in Source: make more
and more sense to me. (Although I'll definitely miss cut and paste when
I'm QA'ing an update package and I already checked out the canonicalness
in the previous version.)
I suppose this is why this was good to continue even though it was not
mandatory. I now have another reasoned out best practice to put to use.
-Toshio
--
Toshio <toshio at tiki-lounge.com>
More information about the fedora-devel-list
mailing list