Messages in this thread | | | Date | Thu, 2 Oct 2008 11:02:46 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH] drivers/serial/8250.c: 'i' may be used uninitialized |
| |
* Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
> On Wed, 1 Oct 2008 01:47:07 -0700 > Steven Noonan <steven@uplinklabs.net> wrote: > > > serial_unlink_irq_chain() does not initialize iterator 'i', and that is > > correct logically because it is always initialized, either in the > > hlist_for_each or in the conditional immediately after (which fires if > > hlist_for_each comes up empty-handed). GCC does not realize this > > connection and emits a false warning. Annotate it with uninitialized_var(). > > > > Signed-off-by: Steven Noonan <steven@uplinklabs.net> > > Ok definitive NAK. gcc 4.3.0 can work this out and doesn't produce a > warning. Thanks for sending the patch though (and to the gcc folks for > rendering it unnecessary)
thanks for sorting it out! A CONFIG_CC_DISABLE_WARNING_ANNOTATIONS=y might be useful as well, which could be used periodically (by gcc folks) to check which warnings are hidden by annotations?
Ingo
| |