Linux Archive

Linux Archive (
-   Fedora Packaging (
-   -   Request for confirmation: Which form is required for a review (

Michael Schwendt 02-08-2012 07:33 PM

Request for confirmation: Which form is required for a review
On Wed, 08 Feb 2012 20:26:06 +0100, MR (Matthias) wrote:

> Hi,
> lately, I stumbled upon a review, which I thought, wouldn't suffice.
> It looks like the following
> name: ok
> summary: ok
> license: ok
> handling locale files: ok
> rpmlint output: only spelling warning
> Not needed BuildRequires: (names), please remove them in git.

Please mention the ticket number.

It may be that the spec file is short/simple and that the listed items
cover most of what was necessary to review. I find it silly to even
mention "name: ok".

> My question is: is this review sufficient,

It is.

There is no requirement for the reviewer to flood the ticket with a huge
list of checkmarks about things that possibly don't even apply to a
package. It doesn't make reviews better, and it doesn't make them safer
either. That is because the guidelines aren't bullet-proof and not
complete either. In other places, the guidelines are not detailed enough
and only experienced reviewers understand the background.

Btw, I think we've had cases before where reviewers, who have posted
an overwhelming checklist, missed several items (or got them wrong).
[not limited to %optflags, plugins in -devel packages, static libs,
licensing, files in wrong subpkgs]

> Do we trust our reviewers,

We do. In the same way we trust our packagers. And still, some packagers
*and* reviewers (re-)introduce packaging mistakes *after* a package has
been approved. ;) Sometimes the changes in package git invalidate the
review results completely, because a packager messes up the packaging.

> so there's no need of bureaucracy?

Fill a growing list of reviews where the reviewer has missed important
items, and then let's figure out what can be done about that. Let's not
punish good reviewers with tiresome bureaucracy.

> Why should/must I do more
> than just setting the flag or writing 7 catchwords?

Do whatever helps you to gain confidence in approving a package. If you
feel it's necessary to process a checklist and include that checklist, do
that. Once you've posted such a list in a review, what would you do if
another reviewer pointed out that you've missed a couple of unowned
directories, for example? (note: that's still a MUST item in the review
packaging mailing list

Michael Schwendt 02-08-2012 08:01 PM

Request for confirmation: Which form is required for a review
On Wed, 08 Feb 2012 21:09:18 +0100, MR (Matthias) wrote:

> I think, a review generates work to do for the reviewer.

That's only one half of the story. :)

First of all, it is the _packager_ who ought to become familiar with the
Packaging Guidelines *and* the Review Guidelines and submit a package that
aims at passing the review process. Then, during the review process, it's
_two_ people (at least) who work on the package as a team and apply best
judgement. It's not only the reviewer. And often, it's the second pair
of eyes that's helpful. Not because the reviewer does "all the work", but
some reviewers spot poor packaging mistakes with the blink of an eye
already, whereas for the package submitter, it may be the 2nd or 3rd
package so far only.

> Why one
> shouldn't see, how much, work or which results were found during this
> work. (even if it reads ... ok, ... ok, ... ).

A checklist may be the result of a quick cut'n'paste job. Mistakes
included. It doesn't tell how much effort (or time) has been spent on the
review. On doing testbuilds, skimming over build logs, examining subpackage
contents and inter-dependencies, and on paying extra attention to details
that perhaps are not covered by the guidelines yet.
packaging mailing list

All times are GMT. The time now is 10:03 PM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.