[cfe-dev] misleading mistake/type in ast matchers documentation

[Samuel Benzaquen]

On Sat, Jan 3, 2015 at 10:19 AM, Manuel Klimek <klimek at google.com> wrote:

>
>
> On Thu Dec 25 2014 at 8:50:46 PM mobi phil <mobi at mobiphil.com> wrote:
>
>> On the page,
>>
>>
>>
>> http://clang.llvm.org/docs/LibASTMatchers.html
>>
>> it is stated
>>
>> "By default, matchers that accept multiple inner matchers use an implicit
>> allOf()
>> <http://clang.llvm.org/docs/LibASTMatchersReference.html#allOf0Anchor>.
>> This allows further narrowing down the match, for example to match all
>> classes that are derived from “Bar”: recordDecl(hasName("Foo"),
>> isDerivedFrom("Bar"))."
>>
>
> This works because isDerivedFrom has an overload that takes a StringRef.
>
>
>> trying that with text based matchers, no result is returned. The correct
>> would be
>>
>> recordDecl(hasName("Foo"), isDerivedFrom(recordDecl(hasName("Bar")))
>>
>> Or the text based matchers do not behave exactly as the equivalent C++
>> matchers?
>>
>> do not care that much about this little mistake in the docu. (if it is a
>> mistake). I care more to know if the string based matchers should behave
>> exactly as the C++ matchers.
>>
>>
> I'm not sure that this is possible (cc'ed Sam); generally, the text based
> AST matchers will always be a simplified version, as they cannot rely on
> all the template magic going on.
>

This case is a simple function overloading, no templates involved. It is
supported in the dynamic matchers.
If isDerivedFrom() didn't accept a string it would give you an error
instead of "0 matches".
The problem here is that matchers created with the macros should not take
StringRef. This is causing a dangling pointer because the string passed to
the matcher constructor is destroyed before the matcher is executed.
Running under ASan shows the problem.

I'll send a fix for it.
Thanks for the report!


>
>
>>
>> rgrds,
>> mobi phil
>>
>> being mobile, but including technology
>> http://mobiphil.com
>>  _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20150104/38549ab8/attachment.html>