[cfe-dev] How to get values from AttributeList
LUIS MIGUEL SANCHEZ GARCIA
luismiguel.sanchez at uc3m.es
Thu Apr 23 08:29:42 PDT 2015
Hi Aaron,
[[own::custom("bb", "aa")]] doesn't work. However, If I add the values by
hand, It works:
Values.push_back(OWNCustom::AA);
Values.push_back(OWNCustom::BB);
return ::new (S.Context) OWNCustomAttr(A.getRange(), S.Context,
Values.data(), Values.size(),
A.getAttributeSpellingListIndex());
Regards,
Luis.
2015-04-22 18:44 GMT+02:00 Aaron Ballman <aaron at aaronballman.com>:
> On Wed, Apr 22, 2015 at 3:10 AM, LUIS MIGUEL SANCHEZ GARCIA
> <luismiguel.sanchez at uc3m.es> wrote:
> > Thanks for your comments!
> >
> > I changed the enumerators like this:
> >
> > def OWNCustom : Attr {
> > let Spellings = [CXX11<"own", "custom">];
> > let Args = [VariadicEnumArgument<"Callable", "Consumed",
> > ["aa", "bb", "cc"],
> > ["AA", "BB", "CC"]>];
> > }
> >
> > however, I still have problems:
> >
> > clang++ test01.cpp -std=c++1y -Xclang -ast-dump -fsyntax-only
> > test01.cpp:8:90: error: use of undeclared identifier 'aa'
> > [[own::custom(bb,aa)]]
> >
> > If I include only one parameter [[own::custom(aa)]], It works.
> >
> > Any idea about what is happening?
>
> It seems that variadic enumerations expect the second (and later)
> arguments to be quoted strings, not identifiers. If you look at
> ParseAttributeArgsCommon, only the first argument can be an
> identifier. I think that's a bug, but in the meantime, you can do:
>
> [[own::custom("bb", "aa")]] instead, and that should work.
>
> ~Aaron
>
> >
> > Regards,
> > Luis.
> >
> >
> > 2015-04-22 0:13 GMT+02:00 Aaron Ballman <aaron at aaronballman.com>:
> >>
> >> On Tue, Apr 21, 2015 at 9:54 AM, LUIS MIGUEL SANCHEZ GARCIA
> >> <luismiguel.sanchez at uc3m.es> wrote:
> >> > Thanks Aaron,
> >> >
> >> > It was the solution. In order to complete the information I include my
> >> > complete example, If anyone needs it.
> >> >
> >> > In the original source code I included an attribute named
> (own::custom)
> >> > that
> >> > receive one or more unsigned int.
> >> >
> >> > [[own::custom(1,2,3,4)]]
> >> > statement;
> >> >
> >> > In file tools/clang/lib/Sema/SemaStmtAttr.cpp, I included:
> >> >
> >> > static Attr *handleOWNCustomAttr(Sema &S, Stmt *St, const
> AttributeList
> >> > &A,
> >> > SourceRange Range) {
> >> > std::vector<unsigned int> Values;
> >> > for (unsigned int i=0;i<A.getNumArgs ();++i){
> >> > Expr *Arg1 = A.getArgAsExpr(i);
> >> > if (auto *L = dyn_cast<IntegerLiteral>(Arg1)) {
> >> > llvm::APInt V = L->getValue();
> >> > unsigned int rint =
> >> > std::atoi(V.toString(10,false).c_str());
> >>
> >> There's no need to use a string to convert the APInt object, btw. ;-)
> >>
> >> > Values.push_back(rint);
> >> > }
> >> > }
> >> > return ::new (S.Context) OWNCustomAttr(A.getRange(), S.Context,
> >> > &Values[0], Values.size(),
> >> >
> >> > A.getAttributeSpellingListIndex());
> >> > }
> >> >
> >> > I hope that it could help anyone.
> >> >
> >> > Now I have another question: I would like to get the same information
> >> > with
> >> > variadic enums as parameters.
> >> >
> >> > I included something like (in
> tools/clang/include/clang/Basic/Attr.td):
> >> >
> >> > def OWNCustom : Attr {
> >> > let Spellings = [CXX11<"own", "custom">];
> >> > let Args = [VariadicEnumArgument<"Callable", "Consumed",
> >> > ["AA", "BB", "CC"],
> >> > ["AA", "BB", "CC"]>];
> >>
> >> You don't want the values and the enumerators to be the same
> >> identifiers; that seems like a dangerous idea.
> >>
> >> > let Documentation = [Undocumented];
> >> > }
> >> >
> >> > and in file tools/clang/lib/Sema/SemaStmtAttr.cpp I included:
> >> >
> >> > static Attr *handleOWNCustomAttr(Sema &S, Stmt *St, const
> AttributeList
> >> > &A,
> >> > SourceRange Range) {
> >> >
> >> > std::vector<OWNCustom::Consumed> Values;
> >> > for (unsigned int i=0;i<A.getNumArgs ();++i){
> >> > IdentifierLoc *Arg1 = A.getArgAsIdent(i);
> >> > OWNCustom::Consumed Out;
> >> > if (OWNCustom::ConvertStrToConsumed (Arg1->Ident->getName(),
> >> > Out)){
> >> > Values.push_back(Out);
> >> > }
> >> > }
> >> >
> >> > return ::new (S.Context) OWNCustomAttr(A.getRange(), S.Context,
> >> > &Values[0], Values.size(),
> >> >
> >> > A.getAttributeSpellingListIndex());
> >> > }
> >> >
> >> > Evething works with only one attribute (e.g. [[own::custom(AA)]]), but
> >> > fails
> >> > with two or more enums (e.g. [[own::custom(AA,BB)]]).
> >> >
> >> > Any help?
> >>
> >> Fails in what way?
> >>
> >> Also, that code has UB when no enumeration values are provided to your
> >> attribute (&Values[0] would be more safely written as Values.data()).
> >>
> >> ~Aaron
> >>
> >> >
> >> > Thanks!
> >> >
> >> >
> >> > Regards,
> >> > Luis.
> >> >
> >> >
> >> > 2015-04-17 17:45 GMT+02:00 Aaron Ballman <aaron at aaronballman.com>:
> >> >>
> >> >> On Fri, Apr 17, 2015 at 11:38 AM, LUIS MIGUEL SANCHEZ GARCIA
> >> >> <luismiguel.sanchez at uc3m.es> wrote:
> >> >> > Hi Aaron,
> >> >> >
> >> >> > I included two integers in attr.td file for this attribute
> >> >> > declaration.
> >> >> > I
> >> >> > included the numbers by hand in the object creation
> >> >> > (I wrote a handler in SemaDeclAttr.cpp file)
> >> >> > and works, but I obtain error when I try to get the values form the
> >> >> > attribute list.
> >> >> >
> >> >> > Do you have any hint?
> >> >>
> >> >> Without seeing your code, it's hard to give a concrete answer, but I
> >> >> suspect what you want is something like:
> >> >>
> >> >> Expr *Arg = A.getArgAsExpr(0);
> >> >> if (auto *L = dyn_cast<IntegerLiteral>(Arg)) {
> >> >> llvm::APInt V = L->getValue();
> >> >> // Do interesting things
> >> >> }
> >> >>
> >> >> ~Aaron
> >> >>
> >> >> >
> >> >> > Thanks!
> >> >> > Luis.
> >> >> >
> >> >> > El 17/04/2015 17:23, "Aaron Ballman" <aaron at aaronballman.com>
> >> >> > escribió:
> >> >> >>
> >> >> >> On Fri, Apr 17, 2015 at 11:05 AM, LUIS MIGUEL SANCHEZ GARCIA
> >> >> >> <luismiguel.sanchez at uc3m.es> wrote:
> >> >> >> > I would like to parser the values of a C++ attribute in clang,
> but
> >> >> >> > I
> >> >> >> > don't
> >> >> >> > know how to get the information from the AttributeList class.
> >> >> >> >
> >> >> >> >
> >> >> >> > For example, is it better to use?
> >> >> >> >
> >> >> >> > IdentifierLoc *Value1 = A.getArgAsIdent(0);
> >> >> >> > IdentifierLoc *Value2 = A.getArgAsIdent(1);
> >> >> >> >
> >> >> >> > or
> >> >> >> >
> >> >> >> > Expr *Value1 = A.getArgAsExpr(0);
> >> >> >> > Expr *Value2 = A.getArgAsExpr(1);
> >> >> >> >
> >> >> >> > And later?
> >> >> >>
> >> >> >> A good place to look is SemaDeclAttr.cpp; that's where parsed
> >> >> >> attributes are converted into semantic attributes that attach to
> >> >> >> declarations. (SemaStmtAttr.cpp is where you'd go for statement
> >> >> >> attributes; type attributes are a bit more involved and live in
> >> >> >> SemaType.cpp).
> >> >> >>
> >> >> >> As for whether you would be using getArgAsIdent or getArgAsExpr,
> >> >> >> that
> >> >> >> depends entirely on the declaration of your attribute in Attr.td
> >> >> >> (specifically, what kind of arguments it takes). Attributes that
> >> >> >> take
> >> >> >> an IdentifierArgument, EnumArgument, or VariadicEnumArgument all
> use
> >> >> >> getArgAsIdent().
> >> >> >>
> >> >> >> HTH!
> >> >> >>
> >> >> >> ~Aaron
> >> >> >>
> >> >> >> >
> >> >> >> > Thanks!
> >> >> >> >
> >> >> >> > Best regards,
> >> >> >> > Luis.
> >> >> >> > --
> >> >> >> > --
> >> >> >> > --
> >> >> >> > --------------------------------------------------
> >> >> >> > Luis Miguel Sánchez García
> >> >> >> > Computer Architecture and Technology Area.
> >> >> >> > Office 2.2.B08
> >> >> >> > Computer Science Department. UNIVERSIDAD CARLOS III DE MADRID
> >> >> >> > Universidad Carlos III de Madrid
> >> >> >> > Avda. de la Universidad, 30
> >> >> >> > 28911 Leganés - Madrid - Spain
> >> >> >> > e-mail: lmsan at arcos.inf.uc3m.es
> >> >> >> > luismiguel.sanchez at uc3m.es
> >> >> >> >
> >> >> >> > Phone: (+34) 91 624 5951
> >> >> >> >
> >> >> >> > Linked-In: http://es.linkedin.com/in/lmsan
> >> >> >> > Twitter: http://twitter.com/lmsanchezgarcia
> >> >> >> > --------------------------------------------------
> >> >> >> >
> >> >> >> > _______________________________________________
> >> >> >> > cfe-dev mailing list
> >> >> >> > cfe-dev at cs.uiuc.edu
> >> >> >> > http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
> >> >> >> >
> >> >
> >> >
> >> >
> >> >
> >> > --
> >> > --
> >> > --
> >> > --------------------------------------------------
> >> > Luis Miguel Sánchez García
> >> > Computer Architecture and Technology Area.
> >> > Office 2.2.B08
> >> > Computer Science Department. UNIVERSIDAD CARLOS III DE MADRID
> >> > Universidad Carlos III de Madrid
> >> > Avda. de la Universidad, 30
> >> > 28911 Leganés - Madrid - Spain
> >> > e-mail: lmsan at arcos.inf.uc3m.es
> >> > luismiguel.sanchez at uc3m.es
> >> >
> >> > Phone: (+34) 91 624 5951
> >> >
> >> > Linked-In: http://es.linkedin.com/in/lmsan
> >> > Twitter: http://twitter.com/lmsanchezgarcia
> >> > --------------------------------------------------
> >
> >
> >
> >
> > --
> > --
> > --
> > --------------------------------------------------
> > Luis Miguel Sánchez García
> > Computer Architecture and Technology Area.
> > Office 2.2.B08
> > Computer Science Department. UNIVERSIDAD CARLOS III DE MADRID
> > Universidad Carlos III de Madrid
> > Avda. de la Universidad, 30
> > 28911 Leganés - Madrid - Spain
> > e-mail: lmsan at arcos.inf.uc3m.es
> > luismiguel.sanchez at uc3m.es
> >
> > Phone: (+34) 91 624 5951
> >
> > Linked-In: http://es.linkedin.com/in/lmsan
> > Twitter: http://twitter.com/lmsanchezgarcia
> > --------------------------------------------------
>
--
--
--
--------------------------------------------------
Luis Miguel Sánchez García
Computer Architecture and Technology Area.
Office 2.2.B08
Computer Science Department. UNIVERSIDAD CARLOS III DE MADRID
Universidad Carlos III de Madrid
Avda. de la Universidad, 30
28911 Leganés - Madrid - Spain
e-mail: lmsan at arcos.inf.uc3m.es
luismiguel.sanchez at uc3m.es
Phone: (+34) 91 624 5951
Linked-In: http://es.linkedin.com/in/lmsan
Twitter: http://twitter.com/lmsanchezgarcia
--------------------------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20150423/bff721db/attachment.html>
More information about the cfe-dev
mailing list