Summary from yesterdays (mini) FESCo meeting

[Michael Schwendt]

On Sat, 30 Dec 2006 18:42:20 +0100, Axel Thimm wrote:

> You imply that the reviewer will just cut'n'paste the list w/o really
> checking the items, or not?

Can you tell that all pasted checkpoints have been checked actually?
No. Not even where you want to require the reviewer to fill in something
meaningful. Example:

  * [...]
  * license: GPL
  * rpmlint is silent
  * BR are complete
  * builds fine for i386
  * haven't tried to build for x86_64
  * debuginfo package not empty
  * file ownership/permissions look good
  * scriptlets look good
  * installs fine
  * seems to start fine
  * removes fine
  * [...]

Ooops. It's a plain copy from the previous approval, where the package was
GPL'ed, too, and the rest identical. For a lot of the checks the result is
the same for all packages. The wording is ambiguous and not detailed
enough. And there is still too much noise. Too much stuff that MUST be
satisfied for an approval anyway. Do you have an idea how many reviews
like that I've done, gpg signed even? Nobody has any interest in that
noise. Packagers don't want to read all that either, especially not if
they are familiar with the guidelines and have created their package
painstakingly. They wait to be told what they need to fix to get approval.

Can you tell what other relevant things have been checked? No, unless
another detailed list is included with the review. If it's not included,
the visual appearance of the review would be reduced to the well-known
check-list.

If you wanted to verify and judge about filled-in values, the protocol
would need to be *much* more verbose and include the diagnostic output
from tools.

  * file ownership/permissions look good
$ rpmls gqview
-rwxr-xrwx  /usr/bin/gqview
-rw-r--r--  /usr/share/applications/gnome-gqview.desktop
drwxr-xrwx  /usr/share/doc/gqview-2.0.4
-rw-r--r--  /usr/share/doc/gqview-2.0.4/COPYING
-rw-r--r--  /usr/share/doc/gqview-2.0.4/ChangeLog
-rw-r--r--  /usr/share/doc/gqview-2.0.4/README
-rw-r--r--  /usr/share/doc/gqview-2.0.4/TODO
drwxr-xrwx  /usr/share/doc/gqview-2.0.4/html
-rw-r--r--  /usr/share/doc/gqview-2.0.4/html/10_1_general.html
Hey, the stuff is world-writable, stupid! ;)

> If you don't then what's all the worry about?

I want to retain the freedom to keep the s/n ratio high by omitting
uninteresting things and being verbose when problems are found or when
hints make sense. I don't want to be forced into a specific format for
approvals of [trivial] packages. If the relationship between packager and
reviewer is fine, it is even possible to do a first brief review of the
most important items and continue in CVS. Similarly, I want to avoid
pedantry. Like 99% of the spec file appear to be American English, but
in Revision 6 the packager added a word with British English spelling.