new check for man pages

[DimStar77]

rpmlint should warn/err on things like this:

/usr/share/man/man1/libcbor.3.gz

i.e. the check should validate that the directory /usr/share/man/man<n> only contains files matching *.<n>%{?ext_man} (ext_man is defined to the compression scheme used, e.g. .gz

[DimStar77]

also: I don't think /usr/share/man/man<n> is supposed to contain further subdirectories; another trigger to yell at packagers

[DimStar77]

Seems we were a bit too strict with the new checks

e.g it is very common for perl modules to have a file called
topic.3pm.gz in /usr/share/man/man3 - which seems to be valid

this implies the folder name man<N> should contain files in the form *.<N>*.gz

(I've seen other suffixed than pm)

[jengelh]

Due to automake, it has been common on GNU/Linux to place .3pm into /usr/share/man/man3.
This is in conflict with the way things were traditionally done (e.g. Solaris) where the full suffix is used, e.g. .1m ../man1m rather than ../man1.
An inquiry on the automake mailing list is pending.

[tmzullinger]

This appears to have caused a regression. Running rpmlint-2.4.0 against https://kojipkgs.fedoraproject.org//packages/shigofumi/0.9/5.fc37/x86_64/shigofumi-0.9-5.fc37.x86_64.rpm results in:

shigofumi.x86_64: W: no-manual-page-for-binary shigofumi

The package contains a matching man page. This works in 2.3.0 or in 2.4.0 if I revert FilesCheck.py to 718e1eb^.

$ rpm -qplv shigofumi-0.9-5.fc37.x86_64.rpm | grep -E ' /usr/(bin|share/man)' | grep -w shigofumi
-rwxr-xr-x    1 root     root                   191240 Jul 23 08:38 /usr/bin/shigofumi
-rw-r--r--    1 root     root                     3941 Jul 23 08:38 /usr/share/man/cs/man1/shigofumi.1.gz
-rw-r--r--    1 root     root                     3372 Aug 18  2015 /usr/share/man/man1/shigofumi.1.gz

Before I go digging, is the cause obvious to anyone here who worked on this change? My first thought is that the translated man page is what the code may be tripping over.

[marxin]

Thanks for the report, apparently I screwed up the regex pattern, I've got a patch candidate.

[tmzullinger]

Thanks @marxin! I was hoping it would be more obvious to someone familiar with the code. I also hope it's an easy fix. Much appreciated!