On Fri, Jan 06, 2006 at 01:13:05PM +0800, Paul Wise wrote: > On Thu, 2006-01-05 at 14:59 -0500, Justin Pryzby wrote: > > > > > > How about the attached combination check - does my check and also does > > > > > yours, with the changes that it checks a couple of other words, and > > > > > checks for a url in the description too. > > > > > > > > > $description =~ m/(homepage|webpage|website)/is > > > > Good, please also add "|URL|upstream" > > > > > > Added, along with some other phrases I found. > > > > What about just 'site' and/or official? (bandwidthd) > > That's caught by the last-paragraph test. Added the phrase "official > site" and a couple others to your test. I'm reluctant to add just "site" > or "official" because I believe they could be fairly common. > > > > If it were an || then any package with a URL in the description would > > > trigger the test, which is probably a bad idea, because not all URLs > > > will be homepages. > > > > My intent was that such packages would use an override; I don't know > > if the numbers are right for this to be feasible. Note that > > I went manually thru most of the packages with urls in their > descriptions, and I didn't find many packages where || helped, mainly it > produced what would be false positives if there was no url in the last > paragraph. > > aptitude -F '%p' search '~d[a-z]+:\/\/[a-zA-Z0-9]' > ~/urls-packages > for f in `cat urls-packages`; do apt-cache show $f >> urls ; done > less urls > Search for: [a-z]+:\/\/[a-zA-Z0-9] > (hold down the down key) > (that can't have been good for my eyes) > > A few problematic packages (that are probably easily fixable): > autoproject beep-media-player-scrobbler bincimap-run bk2site bwbar c2hs > c2hs-doc cbedic cheesetracker cl-ubf clamav-data coriander > daemontools-installer dbtcp dhelp doc-linux-es filler gdesklets-data > gstreamer0.8-ffmpeg gtk2-engines-wonderland gtkatlantic > > Once class of packages that are problematic are those that have the > following at the end of the description (flexloader, galculator). I'm > not sure if these should be dealt with by the homepage test. IMO, > authorship info belongs in debian/copyright: > > Author: Foo Bar <[EMAIL PROTECTED]> > Homepage: http://foo.bar.com > > Another one is packages with multiple homepages (gstreamer0.8-ffmpeg) > > False negatives picked up by using || > > centericq* daemontools-installer dealer efibootmgr freefem3d ghc* gnuvd* > > False positives when using || > > cl-s-xml cupsys-driver-gutenprint doc-linux-nonfree-html exim4 frotz > gaim giftcurs gnat gnome-menus gnuift* greylistd gtk-imonc > > Ultimately, it is up to the lintian maintainers what they want to use in > that test. > > I reviewed packages starting with g and below if you want to review the > rest. > > I also think checks/description needs some tests for proper list > formatting (blank lines before and after, proper indenting). > > > > Of course, all this would be simpler if there was a Homepage field. > > Yes, kind of. Then we would have to encourage people to use it, > > instead of the pseudo-field. So I guess this test can be used for > > that, in the future. :) > > The weird thing is, a few packages appear to have a url field: dict-wn > fastlink fastlink-doc libgtkdatabox-0.2.4.7-0 libgtkdatabox-dev. All of > these appear to be maintained by Andreas Tille <[EMAIL PROTECTED]>. Yea, dpkg allows this, and just outputs during the package building:
dpkg-source: warning: unknown information field `Homepage' in input data in package's section of control info file Also note policy 5.7: "User-defined fields": http://www.debian.org/doc/debian-policy/ch-controlfields.html#s5.7 Thanks again for the patch. -- Clear skies, Justin -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
