Re: Allow gzipped .so and search .so in manpath

[Baptiste Daroussin <bapt@FreeBSD.org>]

[-- Attachment #1: Type: text/plain, Size: 3030 bytes --]

On Thu, Nov 27, 2014 at 01:12:42AM +0100, Ingo Schwarze wrote:
> Hi Baptiste,
> 
> Baptiste Daroussin wrote on Wed, Nov 26, 2014 at 09:24:13PM +0100:
> 
> > I was expecting the above :)
> 
> I just committed the two following patches, and i believe they
> resolve the issue.
> 
> There is still a minor problem in makewhatis(8) with database
> generation when .so to .gz is involved, but i'll get round to that.
> 
> Note that -current is at
> 
>   http://fantadrom.bsd.lv/cgi-bin/cvsweb/
>   CVSROOT=anoncvs@fantadrom.bsd.lv:/cvs
> 
> today, mdocl.bsd.lv is a few days old.  I'm right in the middle of
> switching the server to new hardware.  It will be back at mdocml
> tomorrow, or in a few days at the latest.
> 
> > What about using zlib directly, that could simplify a bunch of things.
> 
> That idea makes sense to me.  However, i want to do the 1.13.2
> release without adding any new dependencies to help the downstream
> projects to stabilize and/or migrate to the 1.13 series.  Granted,
> libz is all but heavyweight, and it's hard to imagine there might
> be anybody out there who doesn't have it, but anyway...
> I'd rather return to that idea after the 1.13.2 release.
> 
> > Right now in the ports tree I'm ripping them at package creation
> > using soelim from groff and I wrote minimalistic version in base
> > to be able to continue using that when groff will be removed.
> 
> Sounds reasonable.
> 
> When dealing with pure .so files, i.e. those containing a single .so
> request and nothing else, it would even be better to replace them with
> hard links instead of having two files with the same content - not so
> much because that saves space, but rather because it allows combining
> information in mandoc.db(5)/apropos(1).
> 
> For the user, the end result looks like this:
> 
> With copies of files:
> 
>    $ apropos getuid
>   geteuid(2) - get user identification
>   getuid(2) - get user identification
> 
> With hard (or soft or .so) links:
> 
>    $ apropos getuid
>   geteuid, getuid(2) - get user identification
> 
> The latter not only saves screen real estate, but also helps to
> realize both are documented in the same manual page.

Yes that makes sense, I'll look in that direction, I ll see what I can cook for
ports.
> 
> Yours,
>   Ingo
> 
Shouldn't the look up for .so links be done in the manpath.

For example I have the WindowMaker manpage (which should really be a link in
that case :D) installed here:

/usr/local/man/man1/WindowMaker.1x.gz

The content is the following:
.so man1/wmaker.1x.gz

If I run:
./mandoc /usr/local/man/man1/WindowMaker.1x.gz

mandoc: man1/wmaker.1x: SYSERR: No such file or directory
mandoc: man1/wmaker.1x:1:19: FATAL: .so request failed: .so man1/wmaker.1x

If I trace it it tries to open man1/wmaker.1x then man1/wmaker.1x.gz but in the
directory where I run the command.

zshall(1) is a better example as in this it cannot be replaced by links.

Best regards,
Bapt

[-- Attachment #2: Type: application/pgp-signature, Size: 181 bytes --]