On 05/31/2017 01:46 PM, Richard Biener wrote: > On Wed, May 31, 2017 at 1:33 PM, Jakub Jelinek <ja...@redhat.com> wrote: >> On Wed, May 31, 2017 at 01:24:47PM +0200, Martin Liška wrote: >>> On 05/31/2017 10:35 AM, Jakub Jelinek wrote: >>>> On Wed, May 31, 2017 at 10:04:53AM +0200, Martin Liška wrote: >>>>> diff --git a/gcc/common.opt b/gcc/common.opt >>>>> index 13305558d2d..5e9942d5100 100644 >>>>> --- a/gcc/common.opt >>>>> +++ b/gcc/common.opt >>>>> @@ -222,9 +222,13 @@ bool flag_opts_finished >>>>> Variable >>>>> unsigned int flag_sanitize >>>>> >>>>> +### >>>>> +Common RejectNegative Joined UInteger Var(flag_no_sanitize_fn) >>>>> PerFunction >>>>> +No sanitize flags for a function >>>> >>>> This looks weird, you are redefining the -### option which is normally >>>> a driver option. >>> >>> I know. I was thinking that it's also a 'dummy' value. >> >> It is not. >> >>>> I would have thought you just want a Variable, like the one right below >>>> this. Aren't all "Variable"s per-function? >>> >>> Unfortunately not. Well, probably adding new type 'PerFunctionVariable' >>> would be >>> solution. Then optc-save-gen.awk needs to be learned how to save/restore >>> these variables. >>> >>> Is it the way we want to go? >> >> Yes. We already have TargetVariable. Or allow specifying >> Variable PerFunction >> >> CCing Joseph as option handling maintainer. > > Just wanting to add that "ab-"using options/variables to implement > what are really > function attributes doesn't look very clean. Unless the plan is to get rid of > function attributes in favor of per-function options.
Well, that was what I did in my original version of the patch. I basically transformed all no_sanitize_address, no_sanitize_undefined and others to a single DECL_ATTRIBUTE called 'no_sanitize' where I masked in integer all these. Feedback I was given by Jakub recommended me to not to do it. > > I'll also note that eventually global variables may want to be no-sanitized > (for asan maybe). And we don't (yet) have per-variable options. Good remark. Martin > > Richard. > >> Jakub
