NetBSD/usr.bin/yacc/NOTES

$NetBSD: NOTES,v 1.3 1997/10/20 03:41:15 lukem Exp $

Berkeley Yacc reflects its origins.  The reason so many routines
use exactly six register variables is that Berkeley Yacc was
developed on a VAX using PCC.  PCC placed at most six variables
in registers.  I went to considerable effort to find which six
variables most belonged in registers.  Changes in machines and
compilers make that effort worthless, perhaps even harmful.

[Given the previous paragraph, and the fact that GCC does not respect
 register declarations, and the fact that much of the rest of the
 4.4lite2 release had "register" declarations extirpated, I've removed
 most of the register declarations from the code. I left them in the
 generated skeleton code "for the hell of it" but they probably should
 go from there, too. -- pm, July 1997]

[I totally removed the register stuff; we've deprecated it elsewhere
 in the tree now -- lukem, October 1997]

The code contains many instances where address calculations are
performed in particular ways to optimize the code for the VAX.
fix .Nm usage, don't geenrate register in skeleton code 1997-10-20 07:41:12 +04:00			$NetBSD: NOTES,v 1.3 1997/10/20 03:41:15 lukem Exp $
Major modernization. 1) register variables have been killed, except for the ones in the generated skeleton (which should probably also be killed soon). 2) __P(())ified prototypes for all functions. 3) new style __RCSID's, and #include <sys/cdefs.h>'s. 4) all warnings generated with WARNS=1 (i.e. all warnings you get from gcc with -Wall -Wstrict-prototypes -Wmissing-prototypes) have been fixed, although not always in the best possible way. in particular on #4, in a couple of places I got "control reaches end of non-void function" errors, and sadly __dead doesn't seem to really work, so I inserted a couple of exit() calls in strategic places. These should be nuked if someone can get __dead to Do The Right Thing. There were also a couple of places where "while (foo = bar)" things cropped up and I didn't do the best conceivable thing, but I usually did. These fixes should probably go back to the byacc maintainers. 1997-07-25 20:46:27 +04:00
update to version 1.9 of Berkeley yacc, from vangogh 1993-05-28 16:30:03 +04:00			`Berkeley Yacc reflects its origins. The reason so many routines`
			`use exactly six register variables is that Berkeley Yacc was`
			`developed on a VAX using PCC. PCC placed at most six variables`
			`in registers. I went to considerable effort to find which six`
			`variables most belonged in registers. Changes in machines and`
			`compilers make that effort worthless, perhaps even harmful.`

Major modernization. 1) register variables have been killed, except for the ones in the generated skeleton (which should probably also be killed soon). 2) __P(())ified prototypes for all functions. 3) new style __RCSID's, and #include <sys/cdefs.h>'s. 4) all warnings generated with WARNS=1 (i.e. all warnings you get from gcc with -Wall -Wstrict-prototypes -Wmissing-prototypes) have been fixed, although not always in the best possible way. in particular on #4, in a couple of places I got "control reaches end of non-void function" errors, and sadly __dead doesn't seem to really work, so I inserted a couple of exit() calls in strategic places. These should be nuked if someone can get __dead to Do The Right Thing. There were also a couple of places where "while (foo = bar)" things cropped up and I didn't do the best conceivable thing, but I usually did. These fixes should probably go back to the byacc maintainers. 1997-07-25 20:46:27 +04:00			`[Given the previous paragraph, and the fact that GCC does not respect`
			`register declarations, and the fact that much of the rest of the`
			`4.4lite2 release had "register" declarations extirpated, I've removed`
			`most of the register declarations from the code. I left them in the`
			`generated skeleton code "for the hell of it" but they probably should`
			`go from there, too. -- pm, July 1997]`

fix .Nm usage, don't geenrate register in skeleton code 1997-10-20 07:41:12 +04:00			`[I totally removed the register stuff; we've deprecated it elsewhere`
			`in the tree now -- lukem, October 1997]`

update to version 1.9 of Berkeley yacc, from vangogh 1993-05-28 16:30:03 +04:00			`The code contains many instances where address calculations are`
			`performed in particular ways to optimize the code for the VAX.`