code review 5485063: 8c: handle 64-bit switch value

8c: handle 64-bit switch value
Cases must still be 32-bit values, but one thing at a time.

[ality, 2011-12-14 02:40:53]

https://codereview.appspot.com/5485063/diff/2001/src/cmd/8c/swt.c
File src/cmd/8c/swt.c (right):

https://codereview.appspot.com/5485063/diff/2001/src/cmd/8c/swt.c#newcode42
src/cmd/8c/swt.c:42: if(n->op != ONAME || n->sym != nodsafe->sym) {
The check for n == nodsafe is a little
suprising but I understand its purpose.
Is that the only way to know if we just
called regsalloc and is it possible for
nodsafe to change inside cgen?

Otherwise, LGTM.

[rsc, 2011-12-14 02:59:37]

> https://codereview.appspot.com/5485063/diff/2001/src/cmd/8c/swt.c#newcode42
> src/cmd/8c/swt.c:42: if(n->op != ONAME || n->sym != nodsafe->sym) {
> The check for n == nodsafe is a little
> suprising but I understand its purpose.
> Is that the only way to know if we just
> called regsalloc and is it possible for
> nodsafe to change inside cgen?

Yes.
You could rename the existing swit1 to swit2
and make swit1 just assume that the reg stuff
is needed and then call swit2.

> Otherwise, LGTM.

Great, now you just need to get this into the
Plan 9 distribution.  :-)

Russ

[rsc, 2011-12-14 05:08:17]

Hello ality@pbrane.org, ken2 (cc: golang-dev@googlegroups.com),

I'd like you to review this change to
https://go.googlecode.com/hg/

[rsc, 2011-12-14 05:08:41]

*** Submitted as http://code.google.com/p/go/source/detail?r=21595dc0395a ***

8c: handle 64-bit switch value
Cases must still be 32-bit values, but one thing at a time.

R=ality, ken2, ken
CC=golang-dev
http://codereview.appspot.com/5485063

[ken3, 2011-12-14 06:11:06]

lgtm