Re: bad change to ginclude/va-alpha.h

["Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>]

 > From: Jeffrey A Law <law@hurl.cygnus.com>
 > 
 >   In message <199808162122.RAA13868@caip.rutgers.edu>you write:
 >   > 	I got a similar error bootstrapping the egcs trunk checked out
 >   > today (19980816) on my alphaev5-dec-osf4.0b box.  I also had the same
 >   > trouble on make other files besides gen-protos.c.  E.g., some of the
 >   > other files which include cpplib.h have this problem, as well as
 >   > cp/errfn.c and cp/tree.c and some files in objc/.
 > 
 > Now what I'm wondering is why you're seeing this and Richard is not.
 > Strange.  I'm also wondering if this is happening on the branch.



	Richard compiled on OSF4.0a, whereas I use 4.0b and Weiwen Liu
used 4.0d.  I don't know what changed in the OS headers between 4.0a and
4.0b which caused this to appear.  I could post the contents of
stdarg.h/varargs.h/stdio.h from my platform if that would help.

	The branch is fine.  This issue arose after installation of
the following patch, which is only in the mainline sources:

 > Sat Aug  1 17:59:30 1998  Richard Henderson  <rth@cygnus.com>
 > 	* ginclude/va-alpha.h (va_list): Use a typedef, not a define.





 >   > 	While I could fix this separately in each file, I'd like to
 >   > fix this once and for all by including stdarg.h/varargs.h in system.h,
 >   > remove it from everywhere else and be done with it.
 > Agreed.
 > 
 >   > 	The following patch takes care of these failures except for
 >   > the objc/ dir since the files making up libobjc.a don't include
 >   > system.h.
 > Looks good.


	Okay, I've install it.  Note the objc dir and possibly more
areas still need to be fixed.  I'll work on that.



 >   > (Should they?  I thought they were to move into a toplevel
 >   > directory? ...)  Anyways, I'll look at doing those in another pass,
 >   > but for now I'd like to include the following patch into the trunk.
 >
 > No, they should not include system.h, or at least some of them shoudn't
 > since they are moving.  Ben just didn't get me the changes in time
 > for the egcs-1.1 branch.
 > Thanks!
 > jeff


	Okay, the objc dir still needs fixing and probably the
libiberty dir too (maybe lib* ?)  I was waiting to hear what the
status of the objc dir was before proceeding with those fixes.
They'll come next.



	Note, a larger issue is that egcs built on OSF4.0[b-d] now has
this problem for *any* sources, not just egcs' own.  Our sources
should be cleaned with respect to this stdarg.h/varargs.h and stdio.h
ordering problem because we should support any stage 1 compiler.

	However we might want to consider having a fix so the 2nd/3rd
stage gcc on OSF4.0 does not have this problem to begin with.  I think
I may need some assistance from Richard in order to understand
what/where to fix though.

		--Kaveh
--
Kaveh R. Ghazi			Project Manager / Custom Development
ghazi@caip.rutgers.edu		Icon CMT Corp.