Re: Conflicting alias for some man pages

[Alejandro Colomar <alx.manpages@gmail.com>]

[-- Attachment #1.1: Type: text/plain, Size: 3807 bytes --]

Hi Helge,

On 12/10/22 08:53, Helge Kreutzmann wrote:
> Hello Branden,
> On Fri, Dec 09, 2022 at 02:37:19PM -0600, G. Branden Robinson wrote:
>> At 2022-12-09T19:53:44+0100, Alejandro Colomar wrote:
>>>>> Could you remove these duplicates in your next upload?
>>>>>
>>>>> I found the following duplicates, I did not do an extensive search:
>>>>> ===================================================================
>>>>> rindex - Both in index.3 and in string.3
>>>>> strncasecmp - Both in strcasecmp.3 and in string.3
>>>>> strncat - Both in strcat.3 and in string.3
>>>>> strncmp - Both in strcmp.3 and in string.3
>>>>> strncpy - Both in strcpy.3 and in string.3
>>>>> __fpurge - Both in fpurge.3 and in stdio_ext.3
>>>>> strcspn - Both in strspn.3 and in string.3
>>>>> strrchr - Both in strchr.3 and in string.3
>>>>> pselect - Both in select.2 and in select_tut.2
>>>
>>> Could you please confirm if this is a bug in the Linux man-pages, or is it
>>> something desirable?
>>
>> I don't think it is a bug for multiple pages to have a mandb entry for
>> the same name.  The man(1) librarian is designed in expectation of that;
>> we have both printf(1) and printf(3), after all.
> 
> Ok. The rationale for my request was that the for *localized* system
> this does not work in Debian (reliably). It only works if the english
> man page is not present, otherwise you get the english version.
> 
> For example, for strcasecmp I currently get the german version,
> however, for strncasecmp I get the english version.
> 
> I reported this to the man-db package in Debian and was told that
> this is a bug in manpages-l10n and that I should create symlinks.
> 
> See https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1020742 for
> details.
> 
> While doing so, I noticed this problem, because how should I decide if
> strncat.3 should point to strcat.3 or string.3? Especially
> automatically, because manpages-l10n has a dozen or so languuages with
> many thousand man pages.
> 
> Currently, the link is set randomly.
> 
> 
>>> I find it a bit weird that we need to specify a NAME only once.
>>
>> There is no such need, and it would be impossible to enforce across
>> projects anyway.
> 
> For me I would report that where I notice it, however, I do see you
> rationale.
> 
> But how should man behave? If you enter
> man strncasecmp
> should it be the man page for strcasecmp.3 or string.3?

The answer is here:

alx@asus5775:~/src/linux/man-pages/man-pages/main$ cat ./man3/strncasecmp.3
.so man3/strcasecmp.3


I don't know how the man pages are preprocessed before they arrive to you, so in 
that processing might be the problem.  Maybe some information is being lost in 
the process.

Cheers,

Alex

> 
>>> Then whatis(1) will not find the other pages that also talk
>>> about an interface (of course, ideally, only a page would describe an
>>> interface, but we know that's not reality).
>>
>> apropos(1) and whatis(1) do indeed behave in a way that surprises me on
>> my Debian system (man-db 2.9.4-2).  I would have expected multiple
>> results.
>>
>> What I expected:
>>
>> $ whatis rindex
>> rindex (3)           - locate character in string
>> string (3)           - string operations
>> [...and maybe others I haven't thought of]
>>
>> What I got:
>> rindex (3)           - locate character in string
>>
>> I am not sure why further matches are being hidden.
> 
> On my Debian testing system this is (more) correct:
> 
> index (3)            - findet ein Zeichen in einer Zeichenkette
> rindex (3)           - locate character in string
> 
> So maybe something on your system? Is this Debian stable or testing?
> 
> Greetings
> 
>           Helge

-- 
<http://www.alejandro-colomar.es/>

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]