Messages in this thread | | | Date | Fri, 8 Sep 2006 16:49:08 -0700 | From | Andrew Morton <> | Subject | Re: [PATCH 10/10] check pr_debug() arguments |
| |
On Fri, 8 Sep 2006 15:55:29 -0700 (PDT) Zach Brown <zach.brown@oracle.com> wrote:
> check pr_debug() arguments > > When DEBUG isn't defined pr_debug() is defined away as an empty macro. By > throwing away the arguments we allow completely incorrect code to build. > > Instead let's make it an empty inline which checks arguments and mark it so gcc > can check the format specification.
Desirable.
> This results in a seemingly insignificant code size increase. A x86-64 > allyesconfig: > > text data bss dec hex filename > 25354768 7191098 4854720 37400586 23ab00a vmlinux.before > 25354945 7191138 4854720 37400803 23ab0e3 vmlinux
Which would indicate that we might have expressions-with-side-effects inside pr_debug() statements somewhere, which is risky. I wonder where?
It looks like the version of gcc which you used is correctly discarding the pr_debug() format string. gcc hasn't always done that, and there's a risk of bloatiness on older gccs. I checked gcc-3.3.2/x86 and it does the right thing, so...
btw, what's up with aio.c using a combination of pr_debug() and dprintk(), and a combination of `#ifdef DEBUG' and `#if DEBUG > 1'? Confusing.
It would be nice to have a single way of doing developer-debug in-tree. We have 182(!) different definitions of dprintk(). Please nobody cc me on that discussion though ;) - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |