<p dir="ltr"><br>
On 29 Apr 2013 20:01, "Aaron Ballman" <<a href="mailto:aaron@aaronballman.com">aaron@aaronballman.com</a>> wrote:<br>
><br>
> Correct; since we're going to be modifying the way this extension<br>
> works, it would be great if we could make it more compatible for<br>
> -fms-compatibility. In this case, MSVC handles void func(xxx) {} when<br>
> /TC is passed in, so it would be good for clang to do so as well<br>
> (along with the other cases that Richard pointed out).</p>
<p dir="ltr">That is unrelated to implicit int in c++, which the current patch addresses, and is in fact unrelated to implicit int at all - Serge confirmed that this is treated as an (ignored) identifier- list, not as a parameter with implicit int.</p>
<p dir="ltr">Unless you have evidence that we need this extension, I don't see any reason to implement it.</p>
<p dir="ltr">> ~Aaron<br>
><br>
> On Mon, Apr 29, 2013 at 1:56 PM, Serge Pavlov <<a href="mailto:sepavloff@gmail.com">sepavloff@gmail.com</a>> wrote:<br>
> > Clang now allows implicit int in C mode (-x c) in some cases:<br>
> ><br>
> > abc = 1;<br>
> > static foo() { return 1; }<br>
> ><br>
> > But in function declaration it is now forbidden unlike to GCC and MSVC:<br>
> ><br>
> > void func(xxx);<br>
> ><br>
> > --Serge<br>
> ><br>
> ><br>
> ><br>
> ><br>
> > 2013/4/30 Aaron Ballman <<a href="mailto:aaron@aaronballman.com">aaron@aaronballman.com</a>><br>
> >><br>
> >> To be clear, MSVC does accept that code so long as you pass in the /TC<br>
> >> flag. So, for instance, the test cases with a .c or .m extension<br>
> >> should allow the implicit int (by default .c files are compiled with<br>
> >> /TC in MSVC), but the .cpp/.mm files should disallow it (unless we<br>
> >> force compilation as C with -x c, which would be the clang version of<br>
> >> /TC).<br>
> >><br>
> >> ~Aaron<br>
> >><br>
> >> On Mon, Apr 29, 2013 at 1:31 PM, Serge Pavlov <<a href="mailto:sepavloff@gmail.com">sepavloff@gmail.com</a>> wrote:<br>
> >> > Hi Richard,<br>
> >> ><br>
> >> ><br>
> >> > 2013/4/29 Richard Smith <<a href="mailto:richard@metafoo.co.uk">richard@metafoo.co.uk</a>><br>
> >> > [...]<br>
> >> >><br>
> >> >> The problem is that clang in C++ mode accepts the code:<br>
> >> >>><br>
> >> >>> int foo(xxx);<br>
> >> >>> Clang intentionally accepts this code due to a check in<br>
> >> >>> Parser::ParseImplicitInt, which appeared in r156854.<br>
> >> >>> The comment in the inserted code states that MS supports implicit int<br>
> >> >>> in<br>
> >> >>> C++ mode, however it looks like none of VS2005, VS2008, VS2010 or<br>
> >> >>> VS2012<br>
> >> >>> does it. So removing the check for MS extension solves the problem.<br>
> >> >><br>
> >> >><br>
> >> >> If it is indeed the case that MSVC does not allow implicit int in C++,<br>
> >> >> then we should absolutely remove that "extension". However, someone<br>
> >> >> presumably added it for a reason, so I'd like to be sure that we've<br>
> >> >> checked<br>
> >> >> this thoroughly before proceeding. Does MSVC allow implicit int in any<br>
> >> >> other<br>
> >> >> contexts? For instance...<br>
> >> ><br>
> >> ><br>
> >> > MSVC doesn't allow implicit int in any context if in C++ mode, details<br>
> >> > are<br>
> >> > in bugzilla.<br>
> >> ><br>
> >> >><br>
> >> >><br>
> >> >> const n = 0; // ok?<br>
> >> >> static f() { // ok?<br>
> >> >> extern m; // ok?<br>
> >> >> return m;<br>
> >> >> }<br>
> >> ><br>
> >> ><br>
> >> > None of these cases are accepted by MSVC.<br>
> >> ><br>
> >> >><br>
> >> >> If MSVC does allow these, then the fix is different: the<br>
> >> >> decl-specifier-seq (or, in C, the declaration-specifiers) for a<br>
> >> >> parameter<br>
> >> >> cannot be empty, so 'int foo(xxx);' would not have implicit int<br>
> >> >> applied,<br>
> >> >> whereas 'int foo(const xxx);' would, and we should make the parser<br>
> >> >> handle<br>
> >> >> that correctly.<br>
> >> >><br>
> >> >>><br>
> >> >>> Another problem - the same code compiled in C mode produces an error,<br>
> >> >>> while both GCC and MSC accept it. To fix it the message<br>
> >> >>> err_ident_list_in_fn_declaration was converted into warning.<br>
> >> >><br>
> >> >><br>
> >> >> Have you checked whether they treat it as an implicit int, or whether<br>
> >> >> they<br>
> >> >> treat it as an (ignored, presumably) identifier list?<br>
> >> ><br>
> >> ><br>
> >> > They are ignored. For instance, both MSVC and GCC successfully compile<br>
> >> > the<br>
> >> > following code:<br>
> >> ><br>
> >> > void abc(xxx);<br>
> >> > void abc(int x, char*y) {}<br>
> >> ><br>
> >> >><br>
> >> >> Also, do you actually have code which relies upon this extension? If<br>
> >> >> not,<br>
> >> >> let's not add it gratuitously.<br>
> >> ><br>
> >> ><br>
> >> > I know nothing about such, the intent was to make behavior more<br>
> >> > compatible.<br>
> >> > Probably it doesn't worth implementation.<br>
> >> ><br>
> >> >> Please split this into its two constituent changes (removing implicit<br>
> >> >> int<br>
> >> >> in microsoft mode, and accepting an identifier-list in a non-defining<br>
> >> >> function declaration). They're basically unrelated, and make more sense<br>
> >> >> to<br>
> >> >> review separately.<br>
> >> ><br>
> >> ><br>
> >> > OK. This patch only removes implicit int in MS-compatibility mode for<br>
> >> > C++.<br>
> >> > Fix to accepting an identifier-list in a non-defining function<br>
> >> > declaration<br>
> >> > is dropped.<br>
> >> ><br>
> >> >>><br>
> >> >>> Files:<br>
> >> >>> include/clang/Basic/DiagnosticSemaKinds.td<br>
> >> >>> lib/Parse/ParseDecl.cpp<br>
> >> >>> lib/Sema/SemaType.cpp<br>
> >> >>> test/Sema/MicrosoftCompatibility.cpp<br>
> >> >>> test/Sema/alloc_size.c<br>
> >> >>> test/Sema/function.c<br>
> >> >>> test/Sema/invalid-decl.c<br>
> >> >>><br>
> >> >>><br>
> >> >>><br>
> >> >>> -----------------------------------------------------------------------------------------------------<br>
> >> >>> diff --git a/include/clang/Basic/DiagnosticSemaKinds.td<br>
> >> >>> b/include/clang/Basic/DiagnosticSemaKinds.td<br>
> >> >>> index 1461716..166dbab 100644<br>
> >> >>> --- a/include/clang/Basic/DiagnosticSemaKinds.td<br>
> >> >>> +++ b/include/clang/Basic/DiagnosticSemaKinds.td<br>
> >> >>> @@ -2314,8 +2314,9 @@ def err_void_only_param : Error<<br>
> >> >>> "'void' must be the first and only parameter if specified">;<br>
> >> >>> def err_void_param_qualified : Error<<br>
> >> >>> "'void' as parameter must not have type qualifiers">;<br>
> >> >>> -def err_ident_list_in_fn_declaration : Error<<br>
> >> >>> - "a parameter list without types is only allowed in a function<br>
> >> >>> definition">;<br>
> >> >>> +def warn_ident_list_in_fn_declaration : Warning<<br>
> >> >>> + "a parameter list without types is only allowed in a function<br>
> >> >>> definition">,<br>
> >> >>> + InGroup<C99Compat>;<br>
> >> >><br>
> >> >><br>
> >> >> This should be an ExtWarn, not a Warning, since this is a required<br>
> >> >> diagnostic per the various C language standards. Also, C99Compat seems<br>
> >> >> wrong.<br>
> >> ><br>
> >> ><br>
> >> > Thank you for the explanation.<br>
> >> ><br>
> >> >><br>
> >> >>><br>
> >> >>> def ext_param_not_declared : Extension<<br>
> >> >>> "parameter %0 was not declared, defaulting to type 'int'">;<br>
> >> >>> def err_param_typedef_of_void : Error<<br>
> >> >>> diff --git a/lib/Parse/ParseDecl.cpp b/lib/Parse/ParseDecl.cpp<br>
> >> >>> index d786ce2..2f0c1a3 100644<br>
> >> >>> --- a/lib/Parse/ParseDecl.cpp<br>
> >> >>> +++ b/lib/Parse/ParseDecl.cpp<br>
> >> >>> @@ -2038,10 +2038,9 @@ bool Parser::ParseImplicitInt(DeclSpec &DS,<br>
> >> >>> CXXScopeSpec *SS,<br>
> >> >>> // error, do lookahead to try to do better recovery. This never<br>
> >> >>> applies<br>
> >> >>> // within a type specifier. Outside of C++, we allow this even if<br>
> >> >>> the<br>
> >> >>> // language doesn't "officially" support implicit int -- we support<br>
> >> >>> - // implicit int as an extension in C99 and C11. Allegedly, MS also<br>
> >> >>> - // supports implicit int in C++ mode.<br>
> >> >>> + // implicit int as an extension in C99 and C11.<br>
> >> >>> if (DSC != DSC_type_specifier && DSC != DSC_trailing &&<br>
> >> >>> - (!getLangOpts().CPlusPlus || getLangOpts().MicrosoftExt) &&<br>
> >> >>> + !getLangOpts().CPlusPlus &&<br>
> >> >><br>
> >> >><br>
> >> >> There is a matching check in lib/Sema/DeclSpec.cpp, and possibly<br>
> >> >> elsewhere. If we're not enabling implicit int in -fms-extensions mode,<br>
> >> >> we<br>
> >> >> need to do that consistently throughout the compiler.<br>
> >> ><br>
> >> ><br>
> >> > Indeed, SemaType.cpp also contains similar check.<br>
> >> ><br>
> >> >>><br>
> >> >>> isValidAfterIdentifierInDeclarator(NextToken())) {<br>
> >> >>> // If this token is valid for implicit int, e.g. "static x = 4",<br>
> >> >>> then<br>
> >> >>> // we just avoid eating the identifier, so it will be parsed as<br>
> >> >>> the<br>
> >> >>> diff --git a/lib/Sema/SemaType.cpp b/lib/Sema/SemaType.cpp<br>
> >> >>> index 8bf5143..243b772 100644<br>
> >> >>> --- a/lib/Sema/SemaType.cpp<br>
> >> >>> +++ b/lib/Sema/SemaType.cpp<br>
> >> >>> @@ -2742,7 +2742,7 @@ static TypeSourceInfo<br>
> >> >>> *GetFullTypeForDeclarator(TypeProcessingState &state,<br>
> >> >>> if (FTI.NumArgs && FTI.ArgInfo[0].Param == 0) {<br>
> >> >>> // C99 6.7.5.3p3: Reject int(x,y,z) when it's not a<br>
> >> >>> function<br>
> >> >>> // definition.<br>
> >> >>> - S.Diag(FTI.ArgInfo[0].IdentLoc,<br>
> >> >>> diag::err_ident_list_in_fn_declaration);<br>
> >> >>> + S.Diag(FTI.ArgInfo[0].IdentLoc,<br>
> >> >>> diag::warn_ident_list_in_fn_declaration);<br>
> >> >>> D.setInvalidType(true);<br>
> >> >><br>
> >> >><br>
> >> >> If you're not issuing an error, you must build a correct AST -- you<br>
> >> >> can't<br>
> >> >> set things invalid.<br>
> >> >><br>
> >> ><br>
> >> > My fault...<br>
> >> > [...]<br>
> >> ><br>
> >> ><br>
> >> > Updated patch:<br>
> >> ><br>
> >> > Files:<br>
> >> > lib/Parse/ParseDecl.cpp<br>
> >> > lib/Sema/DeclSpec.cpp<br>
> >> > lib/Sema/SemaType.cpp<br>
> >> > test/Rewriter/<a href="http://rewrite-byref-in-nested-blocks.mm">rewrite-byref-in-nested-blocks.mm</a><br>
> >> > test/Sema/MicrosoftCompatibility.cpp<br>
> >> ><br>
> >> ><br>
> >> > diff --git a/lib/Parse/ParseDecl.cpp b/lib/Parse/ParseDecl.cpp<br>
> >> > index d786ce2..2f0c1a3 100644<br>
> >> > --- a/lib/Parse/ParseDecl.cpp<br>
> >> > +++ b/lib/Parse/ParseDecl.cpp<br>
> >> > @@ -2038,10 +2038,9 @@ bool Parser::ParseImplicitInt(DeclSpec &DS,<br>
> >> > CXXScopeSpec *SS,<br>
> >> > // error, do lookahead to try to do better recovery. This never<br>
> >> > applies<br>
> >> > // within a type specifier. Outside of C++, we allow this even if the<br>
> >> > // language doesn't "officially" support implicit int -- we support<br>
> >> > - // implicit int as an extension in C99 and C11. Allegedly, MS also<br>
> >> > - // supports implicit int in C++ mode.<br>
> >> > + // implicit int as an extension in C99 and C11.<br>
> >> > if (DSC != DSC_type_specifier && DSC != DSC_trailing &&<br>
> >> > - (!getLangOpts().CPlusPlus || getLangOpts().MicrosoftExt) &&<br>
> >> > + !getLangOpts().CPlusPlus &&<br>
> >> > isValidAfterIdentifierInDeclarator(NextToken())) {<br>
> >> > // If this token is valid for implicit int, e.g. "static x = 4",<br>
> >> > then<br>
> >> > // we just avoid eating the identifier, so it will be parsed as the<br>
> >> > diff --git a/lib/Sema/DeclSpec.cpp b/lib/Sema/DeclSpec.cpp<br>
> >> > index 124f50c..3b3ab2c 100644<br>
> >> > --- a/lib/Sema/DeclSpec.cpp<br>
> >> > +++ b/lib/Sema/DeclSpec.cpp<br>
> >> > @@ -1003,8 +1003,7 @@ void DeclSpec::Finish(DiagnosticsEngine &D,<br>
> >> > Preprocessor &PP) {<br>
> >> > // the type specifier is not optional, but we got 'auto' as a storage<br>
> >> > // class specifier, then assume this is an attempt to use C++0x's<br>
> >> > 'auto'<br>
> >> > // type specifier.<br>
> >> > - // FIXME: Does Microsoft really support implicit int in C++?<br>
> >> > - if (PP.getLangOpts().CPlusPlus && !PP.getLangOpts().MicrosoftExt &&<br>
> >> > + if (PP.getLangOpts().CPlusPlus &&<br>
> >> > TypeSpecType == TST_unspecified && StorageClassSpec == SCS_auto)<br>
> >> > {<br>
> >> > TypeSpecType = TST_auto;<br>
> >> > StorageClassSpec = SCS_unspecified;<br>
> >> > diff --git a/lib/Sema/SemaType.cpp b/lib/Sema/SemaType.cpp<br>
> >> > index 8bf5143..2038f12 100644<br>
> >> > --- a/lib/Sema/SemaType.cpp<br>
> >> > +++ b/lib/Sema/SemaType.cpp<br>
> >> > @@ -793,9 +793,7 @@ static QualType<br>
> >> > ConvertDeclSpecToType(TypeProcessingState &state) {<br>
> >> > // "At least one type specifier shall be given in the declaration<br>
> >> > // specifiers in each declaration, and in the specifier-qualifier<br>
> >> > list in<br>
> >> > // each struct declaration and type name."<br>
> >> > - // FIXME: Does Microsoft really have the implicit int extension<br>
> >> > in<br>
> >> > C++?<br>
> >> > - if (S.getLangOpts().CPlusPlus &&<br>
> >> > - !S.getLangOpts().MicrosoftExt) {<br>
> >> > + if (S.getLangOpts().CPlusPlus) {<br>
> >> > S.Diag(DeclLoc, diag::err_missing_type_specifier)<br>
> >> > << DS.getSourceRange();<br>
> >> ><br>
> >> > diff --git a/test/Rewriter/<a href="http://rewrite-byref-in-nested-blocks.mm">rewrite-byref-in-nested-blocks.mm</a><br>
> >> > b/test/Rewriter/<a href="http://rewrite-byref-in-nested-blocks.mm">rewrite-byref-in-nested-blocks.mm</a><br>
> >> > index 022bb5f..f416b66 100644<br>
> >> > --- a/test/Rewriter/<a href="http://rewrite-byref-in-nested-blocks.mm">rewrite-byref-in-nested-blocks.mm</a><br>
> >> > +++ b/test/Rewriter/<a href="http://rewrite-byref-in-nested-blocks.mm">rewrite-byref-in-nested-blocks.mm</a><br>
> >> > @@ -19,7 +19,7 @@ void f(void (^block)(void));<br>
> >> > - (void)foo {<br>
> >> > __block int kerfluffle;<br>
> >> > // radar 7692183<br>
> >> > - __block x;<br>
> >> > + __block int x;<br>
> >> > f(^{<br>
> >> > f(^{<br>
> >> > y = 42;<br>
> >> ><br>
> >> > diff --git a/test/Sema/MicrosoftCompatibility.cpp<br>
> >> > b/test/Sema/MicrosoftCompatibility.cpp<br>
> >> > new file mode 100644<br>
> >> > index 0000000..15c2558<br>
> >> > --- /dev/null<br>
> >> > +++ b/test/Sema/MicrosoftCompatibility.cpp<br>
> >> > @@ -0,0 +1,4 @@<br>
> >> > +// RUN: %clang_cc1 %s -fsyntax-only -Wno-unused-value -Wmicrosoft<br>
> >> > -verify<br>
> >> > -fms-compatibility<br>
> >> > +<br>
> >> > +// PR15845<br>
> >> > +int foo(xxx); // expected-error{{unknown type name}}<br>
> >> ><br>
> >> ><br>
> >> > --<br>
> >> > Thanks,<br>
> >> > --Serge<br>
> >> ><br>
> >> > _______________________________________________<br>
> >> > cfe-commits mailing list<br>
> >> > <a href="mailto:cfe-commits@cs.uiuc.edu">cfe-commits@cs.uiuc.edu</a><br>
> >> > <a href="http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits">http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits</a><br>
> >> ><br>
> ><br>
> ><br>
</p>