[M3commit] CVS Update: cm3

[Jay Krell]

CVSROOT:	/usr/cvs
Changes by:	jkrell at birch.	10/04/01 17:56:37

Modified files:
	cm3/m3-sys/m3middle/src/: Target.i3 Target.m3 

Log message:
	leave max_align =  64 for all platforms
	this effects:
	I386_FREEBSD / FreeBSD4
	I386_LINUX / LINUXLIBC6
	NetBSD2_i386
	
	It does not affect any others -- not PPC, not Darwin, not 64bit, not Solaris, not NT.
	
	I admit I didn't test the affected platforms.
	
	This is at times overkill, but has some nice properties:
	64bit doubles will be 64 aligned, which is nice, even if not required
	LONGINT will be 64 aligned, which *might* be required for some atomics
	Given that LONGINT is generally implemented as a pair of 32bit integers, this is overkill.
	all platforms have the same alignments
	
	Note, it is tempting to think we can just these values arbitrarily high,
	but we cannot. These values affect interfacing with C.
	If one has
	
	TYPE A = RECORD
	a: INTEGER;
	b: LONGINT;
	END;
	
	on a 32bit system, there may or may not be 4 bytes of padding between
	a and b.
	
	When interfacing with C, it behooves one to always put
	larger fields first, which I believe always avoids/resolves the problem.
	Or if smaller fields precede larger fields, they should be in multiples
	that equal in size what follows them, something like that.
	
	Our interfacing with C is in a few places.
	- m3core/src/unix
	In this case we interface with our own C and put larger fields first
	- m3core/src/win32
	not affected here, merits some review anyway
	- X Windows, merits review
	
	Basically, assuming LONGINT and LONGREAL are absent in any C interfaces,
	i.e. only 32bit types on 32bit platforms,
	this is a no-op. LONGINT is present in m3core/src/unix, but it should be ok.
	LONGINT is "new" so generally absent.
	LONGREAL is a more realistic concern.
	
	To whatever extent Modula-3 is interfaced with Modula-3, we can
	actually set the values arbitrarily high (but it wastes space).