Path: utzoo!utgpu!water!watmath!clyde!bellcore!rutgers!att!pegasus!hansen
From: hansen@pegasus.ATT.COM (Tony L. Hansen)
Newsgroups: comp.lang.c++
Subject: Re: cross-compilation
Summary: what you need to do cross-compilation
Keywords: CC, cross-compilers, c++
Message-ID: <3216@pegasus.ATT.COM>
Date: 25 Oct 88 15:20:07 GMT
References: <5180@saturn.ucsc.edu> <468@otc.oz>
Reply-To: hansen@pegasus.att.com (XT1554000-Tony L. Hansen;LZ 3B-315;3207)
Organization: AT&T Bell Laboratories, Lincroft, NJ
Lines: 38


 > I am trying to cross-compile C++ on a CRAY, [...]
 > This leaves me with the following routines undefined:
 > 	_main 	_new  	_delete

To do C++ (cfront-based) cross-compilation, you will need these things:

    o	a set of C cross-compilation tools, which would include cc, ld, ar,
	nm

    o	a size-alignment table for the remote machine as created by the szal
	program run on that remote machine

    o	a copy of cfront running locally

    o	a version of CC which invokes the cross-compilation versions of cc,
	ld, and nm. It needs to invoke cpp with the options appropriate to
	the remote machine, such as "-U local-machines-symbol" and "-D cray".
	The size-alignment table used should be the one mentioned above.

    o	Using this version of CC, you need to create a local copy libC.a.
	Compile all of the .c files into .o files, using the -c option to CC,
	and use the cross-compilation version of ar to create libC.a. If you
	don't have a version of ar, ship the .o's across to the remote
	machine, do the ar there, and ship the libC.a back.

    o	Now modify CC again so that it points to the cross-compilation
	version of libC.a.

    o	You might have to modify the munch/patch line to work with your
	systems output from nm.

The new version of CC should now be usable for doing a complete
cross-compilation of C++ programs. I've used the above technique for doing
cross-compilation to several other systems.

					Tony Hansen
				att!pegasus!hansen, attmail!tony