[cfe-commits] [PATCH] Thread-safety analysis: downgrade lockable requirement from error to warning.

[Richard Smith]

LGTM; one trivial comment added to codereview.

I also have some suggestions for diagnostic wording, but they're unrelated
to the patch itself:
def warn_attribute_argument_not_lockable : Warning< "%0 attribute requires
arguments whose type is annotated " "with 'lockable' attribute">,
InGroup<ThreadSafety>,
DefaultIgnore;

It would be useful for this diagnostic to point at the problematic argument
and mention its type: "'guarded_by' attribute argument type 'MyMutex' is
not annotated with 'lockable' attribute"

def warn_attribute_argument_not_class : Warning< "%0 attribute requires
arguments that are class type or point to class type">, InGroup<ThreadSafety>,
DefaultIgnore;
This would read more naturally as "requires arguments that are of class
type or pointer to class type". It would be useful to include the type here
too.

On Thu, Mar 15, 2012 at 9:26 AM, Delesley Hutchins <delesley at google.com>wrote:

> This patch downgrades the requirement that mutex expressions must
> refer to lockable objects from an error message to a warning.  The
> rest of the thread safety analysis works fine even if the class is not
> lockable, so there is no reason to break the build just because a
> class is missing the LOCKABLE attribute.
>
> http://codereview.appspot.com/5820063/
>
>  -DeLesley
>
> --
> DeLesley Hutchins | Software Engineer | delesley at google.com | 505-206-0315
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20120321/27d5b8db/attachment.html>