Re: LYNX-DEV autoconfig changes for SCO OpenServer

[Bela Lubkin]

Larry W. Virden wrote:

> The -R flag should be relevant to all ELF based systems (System V.4 based).
> It tells the linker to place the following argument (a directory) into
> the .so in the run time dynamic link resolver's PATH.  In this way,
> the executable can find all the .so files needed for runtime resolution of
> the program.

"should be relevant", perhaps, but it is not.  The OpenServer toolchain
is based on, but not identical, to the UnixWare 2.0 toolchain.  It has a
-R flag with a completely different meaning.  Also, the UnixWare 2.0 and
2.1 ld's have no -R flag at all.  I think this is in fact relevant only
to Solaris 2.x (or perhaps the GNU ld, which is I believe what's used in
Linux and BSD, has also followed suit?)

Now, you tell me, is there any relevance whatever to a pair of flags "-L
foo -R foo" -- pointing both to the same directory?  Or does the linker
automatically assume -R foo based on -L foo?  In any case, isn't
/usr/ccs/lib the default search path for the runtime linker, and if so,
what is the point of specifying it?

T.E.Dickey wrote:

> but -- what's a good rule for determining if it's applicable?

That depends on the answer to my questions of Larry.  If the -R flag
being used in the makefile is completely irrelevant, then just remove
that whole clause of configure.  If it isn't irrelevant, do what
configure usually does: compile a test program.  If you can `cc
-L/usr/ccs/lib -R/usr/ccs/lib` a program, then run it successfully, you
at know that it is at least not immediately fatal to specify those
flags...

>Bela<
;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;