Hi -
nathans wrote:
> [...]
> Fast forward several months and sure enough, we have an outdated copy
> of Vector in the rpms (v1.0.2 tagged >2 months ago), noone has updated
> the second webjs copy in a timely fashion, and noone is pushing latest
> Vector code into rawhide.
What's missing here is a problem statement. There's no obligation we
always ship the latest release. Which of the 1.0.1-to-1.0.2 changes
did you judge as relevant & useful? We provide a service to our users
in judging, packaging, testing, integrating (top level index.html!).
In this case, that meant skipping that release.
> > How well tested is it?
>
> Works nicely for me, tested it a fair bit yesterday to double check
> but I think trusting Martin's when-to-release judgement is the best
> strategy for the RPMs we build.
>
> > e.g., on F21 Firefox it doesn't run well.
> > https://github.com/Netflix/vector/issues/83
>
> I'm sure it will get resolved soon [...]
Indeed it is, in vector git master.
But one can't claim at the same time that one should automatically
trust an upstream release, and that token testing is sufficent, when
such a showstopper bug is in our face (as you must agree, since you
bundled an unversioned 1.0.1 vector tarball into the new fedora 3.10.7
rpms).
> [...] (and sooner on rawhide, via the master branch pull in
> scripts/spin-rawhide now).
Well, maybe. The way this part of your commit works, the script just
uses vector tarball name pulled out of the spec Sources:. So one has
to take an affirmative step in hand-editing that line out of the spec
file to trigger the new git clone from github/vector. That in turn
generates a tarball that is not comparable to the released ones. Not
just for the farabovementioned reasons (not being a
"built"/compressed/minified dist/), but it's not even complete &
executable. Did you test it?
That's not to say that a git/master vector along with a git/master pcp
would be undesirable in rawhide. Just that this is not what the new
spin-rawhide script does. (And for a real pcp+webjs release, one
would normally prefer a good stable release for both rather than a
snapshot, which brings us back to manual judgement & testing & c.)
- FChE
|