Re: [Update REQ 5660]: Distribution and Makefiles

[rp2y+@CS.cmu.edu]

Subject: Re: [Update REQ 5660]: Distribution and Makefiles 
Date: Wed, 29 Nov 1995 19:52:23 -0500

> From: Stuart Levy <slevy>
> Date: Wed, 29 Nov 1995 15:55:50 -0600
> Message-Id: <199511292155.PAA04553 at royden.geom.umn.edu>
> To: software@geom.umn.edu, software@geom.umn.edu
> Subject: Re:  [ REQ 5660]: Distribution and Makefiles
> Cc: (software-admin at geom.umn.edu) (software-news at geom.umn.edu)
> 
> Interesting -- could you explain what's wrong with creating hard links
> into one's own parent directory?

Nothing, if the parent directory is on the same device...  The way AFS does 
things is that it (apparently) treats all directories as if they reside on 
different devices (as they sometimes do).  I'm sure there is a good reason for 
this, although I'm not quite sure what it is (could be something to with the 
way it does caching and/or replication).

Obviously, the tar file can be dealt with relatively easily (using the L 
option).  The real problem is the lack of a global LN definition.

> Yes, gmake's complaints about multiple references to the same dependency in
> a rule are annoying, but as far as I can tell they're completely harmless --
> it seems to build anyway, when I do it.  Do you find otherwise?

No, I finally gave up on changing your Makefiles (there's just too damn many 
of them...) and untarred the source tree on local disk and did the compile 
(everything went fine once I figured out that the definition of RANLIB in 
mk.alpha was ":" rather than "ranlib").

Thanks for the response,
Robert