no (or bad) support for diverted or conflicting man pages

[Scott Moser]

I'm guessing its a very similar root cause, to bug 336328 or bug 544332.

http://manpages.ubuntu.com/manpages/lucid/man5/fstab.5.html
shows the man page is from loop-aes-utils_2.15.1~rc1-2ubuntu1_i386 .

More likely, on a user's system, 'mount' provides /usr/share/man/man5/fstab.5.gz

In the debs, both loop-aes-utils and mount provide fstab.5.gz. loop-aes-utils diverts /usr/share/man/man5/fstab.5.gz to /usr/share/man/man5/fstab-orig.5.gz

In the processing of the archive, loop-aes-utils man page "won", although it is less complete in some respects.

One potential work around for this (and possibly the other bugs) would be to collect all versions of a man page, and if there was more than one, provide links to the specific ones.
Basically, providing per-package man pages if a collision occured.

[Dustin Kirkland  (kirkland)]

Confirmed, known, long standing issue with no simple solution.

Ideally, the package installing the conflicting manpage would respect the global namespace, and add an iterator within its manpage file naming.

Search manpages.ubuntu.com for "open":
 * http://manpages.ubuntu.com/cgi-bin/search.py?cx=003883529982892832976:5zl6o8w6f0s&cof=FORID:9&ie=UTF-8&q=open&titles=Title&lr=lang_en&siteurl=manpages.ubuntu.com/manpages/lucid/en/man3/open.3posix.html

There's a section 3 for open, for perl, tcl, and posix.

That's much cleaner.

[Clint Byrum (clint-fewbar)]

The iterator is helpful and definitely a best practice. we should report bugs on any packages that don't respect that namespace.

However, even in that situation, its not clear how to link to the output of 'man 3 open' that one would expect on a default or common install.

This one is particularly troubling:

http://manpages.ubuntu.com/manpages/man5/init.5.html

This is particularly frustrating as the best practice on learning about upstart is 'read man 5 init' .. but if you try to use the web to do it, you get runit-run, which is an init replacement from universe.

Seems that one simple thing to do would be to make sure the repositories are generated in least important -> most. So at least do main *last* so that its man pages will overwrite those of any other repository.

That can be fixed by switching the default order that is used right now.

Longer term, I'm testing a fix right now that builds an index file of conflicting pages so you get the Provided by's with links to each page.