Missing braces in Updates.h macro

[Donald Bruce Stewart]

dons:
> dons:
> > Hey,
> > 
> > As you might have seen, last night's head on OpenBSD failed with:
> > 
> >         ../../ghc/compiler/ghc-inplace -optc-O -optc-Wall -optc-W -optc-Wstrict-prototypes -optc-Wmissing-prototypes -optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return -optc-Wbad-function-cast -optc-I../includes -optc-I. -optc-Iparallel -optc-DCOMPILING_RTS -optc-fomit-frame-pointer -H16m -O  -O2 -static -#include Prelude.h -#include Rts.h -#include RtsFlags.h -#include RtsUtils.h -#include StgRun.h -#include Schedule.h -#include Printer.h -#include Sanity.h -#include Storage.h -#include SchedAPI.h -#include Timer.h -#include Itimer.h -#include ProfHeap.h -#include LdvProfile.h -#include Profiling.h -#include Apply.h -fvia-C -dcmm-lint     -c Interpreter.c -o Interpreter.o
> >         Interpreter.c: In function `interpretBCO':
> >         Interpreter.c:405: syntax error before `const'
> >         Interpreter.c:405: `info' undeclared (first use in this function)
> >         Interpreter.c:405: (Each undeclared identifier is reported only once
> >         Interpreter.c:405: for each function it appears in.)
> >         gmake: *** [Interpreter.o] Error 1
> > 
> > I tracked this down, and it looks like it is barfing on the occurence of
> > DECLARE_IPTR(info) in UPD_REAL_IND. A similar situation occurs in
> > Schedule.c:3218: "syntax error before `const'". Looks like it is because
> > of the decl not appearing at the start of a block. 
> > 
> > It is still a bit mysterious -- as the head is building fine on linux
> > each night. Any gcc hackers know why this could be? I'm using 2.95.3
> > here.
> > 
> > The following patch makes these cases compile.
> > 
> > -- Don
> >         
> > ------------------------------------------------------------------------
> > --- Updates.h.orig	Thu Sep  2 11:25:31 2004
> > +++ Updates.h	Thu Sep  2 11:25:36 2004
> > @@ -57,13 +57,15 @@
> >     if you *really* need an IND use UPD_REAL_IND
> >   */
> >  #define UPD_REAL_IND(updclosure, ind_info, heapptr, and_then)		\
> > +	{ 								\
> >         DECLARE_IPTR(info);						\
> >         info = GET_INFO(updclosure);					\
> >          AWAKEN_BQ(info,updclosure);					\
> >         updateWithIndirection(GET_INFO(updclosure), ind_info,		\
> >                               updclosure,				\
> >                               heapptr,					\
> > -			      and_then);
> > +			      and_then); 				\
> > +	}
> 
> Ah, but it looks like adding braces to the macro then breaks its use Cmm code:
> 
>         ../../ghc/compiler/ghc-inplace -H16m -O  -O2 -static -#include Prelude.h -#include Rts.h -#include RtsFlags.h -#include RtsUtils.h -#include StgRun.h -#include Schedule.h -#include Printer.h -#include Sanity.h -#include Storage.h -#include SchedAPI.h -#include Timer.h -#include Itimer.h -#include ProfHeap.h -#include LdvProfile.h -#include Profiling.h -#include Apply.h -fvia-C -dcmm-lint     -c Updates.cmm -keep-tmp-files -o Updates.o
>         Updates.cmm:61:160: parse error on input `{'
> 
> This is occuring in the use of UPD_REAL_IND, by UPD_SPEC_IND in
> UPD_FRAME_ENTRY_TEMPLATE, in Updates.cmm. 
> 
> Maybe Simon has a better idea what is going on here?


Explicitly wrapping braces around the two uses of UPD_IND in C code that
don't compile let the entire rts go through. Yay. Patch below. 

Still don't know why this isn't happening on the nightly build machines.

-- Don

------------------------------------------------------------------------
Index: Interpreter.c
===================================================================
RCS file: /cvs/fptools/ghc/rts/Interpreter.c,v
retrieving revision 1.47
diff -u -u -r1.47 Interpreter.c
--- Interpreter.c	13 Aug 2004 13:10:02 -0000	1.47
+++ Interpreter.c	2 Sep 2004 04:09:08 -0000
@@ -402,7 +402,9 @@
 	// Returning to an update frame: do the update, pop the update
 	// frame, and continue with the next stack frame.
 	INTERP_TICK(it_retto_UPDATE);
+        {
 	UPD_IND(((StgUpdateFrame *)Sp)->updatee, obj); 
+        }
 	Sp += sizeofW(StgUpdateFrame);
 	goto do_return;
 
Index: Schedule.c
===================================================================
RCS file: /cvs/fptools/ghc/rts/Schedule.c,v
retrieving revision 1.201
diff -u -u -r1.201 Schedule.c
--- Schedule.c	19 Aug 2004 11:27:45 -0000	1.201
+++ Schedule.c	2 Sep 2004 04:09:09 -0000
@@ -3215,7 +3215,9 @@
 		SET_HDR(raise_closure, &stg_raise_info, CCCS);
 		raise_closure->payload[0] = exception;
 	    }
+            {
 	    UPD_IND(((StgUpdateFrame *)p)->updatee,raise_closure);
+            }
 	    p = next;
 	    continue;