[PATCH] D56571: [RFC prototype] Implementation of asm-goto support in clang

[Eli Friedman via Phabricator via cfe-commits]

efriedma added inline comments.


================
Comment at: lib/Sema/JumpDiagnostics.cpp:347
+      LabelAndGotoScopes[S] = ParentScope;
+      Jumps.push_back(S);
+    }
----------------
jyu2 wrote:
> efriedma wrote:
> > This doesn't look right; I think we need to add it to IndirectJumps instead.  This probably impacts a testcase like the following:
> > 
> > ```
> > struct S { ~S(); };
> > int f() {
> >   {
> >     S s;
> >     asm goto(""::::BAR);
> >     return 1;
> >   }
> > BAR:
> >   return 0;
> > }
> > ```
> > 
> > (gcc currently accepts this and skips running the destructor, but I'm pretty sure that's a bug.)
> Hi Eli,
> I see both g++ and clang++ with my change call ~S.  Am I missing something?  Thanks.
> 
> 1>clang++ j.cpp
> 1>./a.out
> ~S()
> 1>g++ j.cpp
> 1>./a.out
> ~S()
> 
> Here is the test case:
> 1>cat j.cpp
> extern "C" int printf (const char *,...);
> struct S { ~S() {printf("~S()\n");}; };
> int f() {
>   {
>     S s;
>     asm goto(""::::BAR);
>     return 1;
>   }
> BAR:
>   return 0;
> }
> 
> int main()
> {
>   f();
> }
> 
Oh, sorry, I wasn't paying attention to the actual contents of the asm.  If you modify the asm goto slightly, to `asm goto("jmp %0"::::BAR);`, you'll see that the destructor doesn't run, at least with gcc.  (This is different from the way `goto BAR;` works.)


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D56571/new/

https://reviews.llvm.org/D56571