This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
On Tue, Mar 20, 2007 at 05:32:59PM +0100, Dirk Mueller wrote: > On Tuesday, 20. March 2007, Richard Guenther wrote: > > > I still would prefer you'd warn at dereference or escape sites only > > (escape sites > > are possible dereference sites). > > That would disable warnings about taking addresses of array references that > are out of bounds. Right?
If you never dereference such a pointer and it doesn't escape.
char a[10]; char *r;
int foo (void) { char *p = &a[12]; // This wouldn't warn char *q = &a[12]; *q = 5; // We would warn here r = &a[-1]; // and here as well bar (); }
But if you never dereference it and it doesn't escape, then it is completely harmless.
Yes, and in most cases (as above) the address taking would be removed by DCE anyway.
As is -Warray-bounds just causes too many false positives, see PR31227 or e.g. yesterday when I tried to bootstrap with obj-c++ among --enable-languages trunk didn't bootstrap because of it either.
Jakub
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |