Subject: Re: RFC: Fix for IMAKE_MANINSTALL handling
To: Roland Illig <rillig@NetBSD.org>
From: Johnny C. Lam <email@example.com>
Date: 10/31/2007 17:48:42
Roland Illig wrote:
> Joerg Sonnenberger wrote:
>> Let me repeat: currently imake installs either cat pages or man pages. I
>> don't know of a platform where it does both. Which platform do you have
>> in mind for that?
> My computer. It has NetBSD 4.99.30, with X11 replaced by XFree86
> 4..0, as provided by
Please forgive my late entry into this discussion. I see some commits
have started to trickle into pkgsrc, so I'll offer an implementation
suggestion. I'd prefer to keep all the hard stuff in mk/plist if
possible, and leave PLISTs and simple to read and understand as possible.
In the above scenario, where you have a custom native X11 installation,
you say that both man pages and catman pages are installed. In your
oneko example, why don't we do something like:
We remove all those IMAKE_* variables from mk/platform and instead we
compute them from the installed imake *.cf files. We also set
IMAKE_MANINSTALL appropriately (either "maninstall", "catinstall", or
"maninstall catinstall") depending on what we see in the *.cf files. If
IMAKE_MANINSTALL is "maninstall catinstall", then for each man page
entry in the PLIST, we insert a corresponding catman page entry.
In the case where a package installs manpages through imake, we do
where IMAKE_MANDIR varies based on the X11 installation. This would
inform the plist module that the "man" subdirectory could be something
else, e.g. "catman/u_man" on IRIX. We teach the plist module a new
variable, PKGMANSUBDIR which is a special directory inside the normal
man page directory where the man pages are stored, e.g. "X11" on IRIX
but normally empty for most platforms.
Then the plist module would produce the following on my vanilla
and the following on IRIX:
and the following on your system:
Would this work? Are there any possible problems with this approach?
-- Johnny C. Lam