Re: Solaris Binary Installation Report/Gripe

[Douglas Orr]

At 08:10 AM 9/10/97 -0600, glen e. p. ropella wrote:
>
>I'll defer to Doug for the explanation of why we're using a
>big object file as opposed to a randomized archive.  And
>we have absolutely no excuse for not having a better README.
>But, I do want to address some stuff.

The problem we're solving with the binary distribution (and, there are
several we're not solving, as you've noticed) is that it's nearly
impossible to get the right combination of versions of the eight or twelve
odd component libraries Swarm uses.  So, I "solve" the problem by using the
swarm components linked up to induce ld to extract the pieces of
tk/tcl/whateverelse it may need from a set of archives we happen to know
has exactly the right combination of .o's within.  

The side effect of this process of discovering what we really need within
tk/tcl/... is, the resulting object file gets all the correct library
.o's... and all of the swarm components baked into it as well.  Which
brings us to Barry's problem.  Once the swarm components are in, most
linkers don't know how to get them out.

Now, as to whether there are any ways to not get them in, in the first
place: I hadn't thought of it before, but I *might* be able to do something
like, figure out what components would be used from the archives by using
something like lorder and tsort, the way we used to build properly ordered
archives in the good-old-days before the invention of ranlib.  But, given
that these are the good-new-days after the invention of ranlib, I wouldn't
totally guarantee that either lorder or tsort are available or in working
order on all of our platforms, given that who in their right mind would use
them now, present company excepted.  And, I don't remember if you have
exactly the information you need from those programs to avoid .o's that
aren't referenced.

But, as I say, it's a possibility.  And, if someone else has some more
obvious solution to extracting only the archive members you would use in a
link, but not linking them, I'll happily take a look at any other
reasonable suggestion...

        -Doug

>
>Barry McMullin writes:
> > in the "standard" places.  I knew this *was* a standard
> > problem, so I checked the FAQ but couldn't find
> > anything there; tried search on the Swarm site for
>
>Well, I apologize for the web-site not pointing to the 
>latest FAQ.  But, it has been mentioned in the Pheromone
>and on the Swarm-Support list that Paul Johnson has stepped
>up to redo the FAQ.  The latest version is on his site:
>
>http://lark.cc.ukans.edu/~pauljohn/linux/FAQ.html
>
>This will soon be remedied.  I am going to completely upgrade
>the docs and the web pages for this upcoming release (1.0.3,
>which will be ready any minute now [grin]).
>
> > conflict).  So I suggest this issue should be added to
> > the FAQ, and the README (at least in the binary
> > distributions).
>
>Yes, you're right.  And they will be.
>
> > o Why are there no 1.0.2 binaries?
>
>Mainly because I've been too lazy to build them.  Actually,
>I've been badgered by several people about this.  The 
>primary reason I haven't done it yet is because I was 
>hoping to get SGI and DEC Alpha binaries.  When I got those,
>I was going to update the entire collection.  I never got
>those, and so the entire process broke down.  [sigh]  And
>I understand that that's not sufficient excuse.  There
>will be 1.0.3 binaries.
>
> > Having said that, I want to end by saying that I still
> > very much support the idea of binary distributions, and
> > they really can make swarm significantly more
> > accessible.
>
>Agreed!
>
>glen
>-- 
>{glen e. p. ropella <address@hidden> |  Send lawyers, guns, and money!  }
>{Hive Drone, SFI Swarm Project         |            Hail Eris!            }
>{http://www.trail.com/~gepr/home.html  |               =><=               }
>
>                  ==================================
>   Swarm-Support is for discussion of the technical details of the day
>   to day usage of Swarm.  For list administration needs (esp.
>   [un]subscribing), please send a message to <address@hidden>
>   with "help" in the body of the message.
>                  ==================================
>
>

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ==================================