code review 7780044: cmd/gc: silence valgrind error

cmd/gc: silence valgrind error

valgrind complained that under some circumstances, 

    *nr = *nc

was being called when nr and nc were the same *Node. The suggestion my Rémy was to introduce a tmp node to avoid the potential for aliasing in subnode.

[dave_cheney.net, 2013-03-13 10:58:52]

Hello remyoudompheng@gmail.com, minux.ma@gmail.com, rsc@golang.org (cc:
golang-dev@googlegroups.com),

I'd like you to review this change to
https://code.google.com/p/go

[dave_cheney.net, 2013-03-13 10:59:42]

The original valgrind report.

==2499== Memcheck, a memory error detector
==2499== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
==2499== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright info
==2499== Command: /home/dfc/go/pkg/tool/linux_386/8g -o
/tmp/go-build072428647/fmt/_obj/_go_.8 -p fmt -complete -D
_/home/dfc/go/src/pkg/fmt -I /tmp/go-build072428647
/home/dfc/go/src/pkg/fmt/doc.go /home/dfc/go/src/pkg/fmt/format.go
/home/dfc/go/src/pkg/fmt/print.go /home/dfc/go/src/pkg/fmt/scan.go
==2499== Parent PID: 2302
==2499==
==2499== Source and destination overlap in memcpy(0xbefc76d8, 0xbefc76d8, 224)
==2499==    at 0x402D9A9: memcpy (in
/usr/lib/valgrind/vgpreload_memcheck-x86-linux.so)
==2499==    by 0x8049071: subnode (cplx.c:364)
==2499==    by 0x8048F11: complexmove (cplx.c:70)
==2499==    by 0x8055ED7: gmove (gsubr.c:1288)
==2499==    by 0x804ADCE: cgen (cgen.c:125)
==2499==    by 0x806AB3D: cgen_as (gen.c:745)
==2499==    by 0x80503A9: cgen_callret (ggen.c:316)
==2499==    by 0x804B5DB: cgen (cgen.c:366)
==2499==    by 0x806AB3D: cgen_as (gen.c:745)
==2499==    by 0x806A72F: gen (gen.c:468)
==2499==    by 0x806A0C9: genlist (gen.c:257)
==2499==    by 0x80501D8: cgen_call (ggen.c:259)

[rsc, 2013-03-13 16:48:29]

LGTM

Thank you. I have yet to see a valgrind report that was a false positive.
This might be the closest one I've seen - who knew that the compiler was
putting a memcpy there! - but it's still worth cleaning up to make the rest
of the valgrind output useful.

Russ

[dave_cheney.net, 2013-03-13 19:42:14]

Excellent, thank you. I'm working through all the permutations of
{5,6,8}{a,c,g,l}. 



On 14/03/2013, at 3:48 AM, Russ Cox <rsc@golang.org> wrote:

> LGTM
> 
> Thank you. I have yet to see a valgrind report that was a false positive. This
might be the closest one I've seen - who knew that the compiler was putting a
memcpy there! - but it's still worth cleaning up to make the rest of the
valgrind output useful.
> 
> Russ

[dave_cheney.net, 2013-03-13 20:11:26]

*** Submitted as https://code.google.com/p/go/source/detail?r=28dbe614d61c ***

cmd/gc: silence valgrind error

valgrind complained that under some circumstances, 

    *nr = *nc

was being called when nr and nc were the same *Node. The suggestion my Rémy was
to introduce a tmp node to avoid the potential for aliasing in subnode.

R=remyoudompheng, minux.ma, rsc
CC=golang-dev
https://codereview.appspot.com/7780044