Re: PATCH multi-arch GET_LONGJMP_TARGET

[Richard Earnshaw <rearnsha@arm.com>]

[-- Attachment #1: Type: text/plain, Size: 1467 bytes --]

> > 
> > This patch makes that
> > GET_LONGJMP_TARGET is always defined for all mutiarched target
> > (even partial !).
> > 
> > Thus it breaks compilation of Cygwin native GDB,
> > because JB_PC and JB_ELEMENT_SIZE are not defined for these targets.
> > (and several other i386 targets)
> > 
> > The correct fix would be to add these values to gdbarch ...
> > 
> > The following was committed as an obvious fix.
> 
> Ah, sorry about that.
> 
> Perhaps it would also be sensible to check GET_LONGJMP_TARGET_P() in that 
> code and always return 0 if it is false.
> 
> I'll go around the other tdep files and clean those up as well.
> 
> R.
> 

Thinking about it, the cleanest way to handle this for now is to change 
the test for a definition of JB_PC, since unless that is defined the code 
won't compile, and even if it did it would give the wrong answer.

As far as I can tell, apart from i386, only two further ports define 
GET_LONGJMP_TARGET: arc and sparc.  Arc seems never to use the code 
(there's only one .h file for that port and it doesn't define 
GET_LONGJMP_TARGET or JB_PC); Sparc seems to define it for pretty much all 
configurations I can think of.

Anyone object to the following patch?  I'll leave the x86 code as is for 
now, since you've already fixed that.

R.

<date>  Richard Earnshaw  <rearnsha@arm.com>

	* arc-tdep.c (get_longjmp_target): Only compile this function if
	JB_PC is defined.
	* sparc-tdep.c (get_longjmp_target): Likewise.



[-- Attachment #2: gdb-jbpc.patch --]
[-- Type: text/x-patch , Size: 1918 bytes --]

Index: arc-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/arc-tdep.c,v
retrieving revision 1.8
diff -p -r1.8 arc-tdep.c
*** arc-tdep.c	2002/02/05 04:37:21	1.8
--- arc-tdep.c	2002/02/19 10:35:26
*************** arc_software_single_step (enum target_si
*** 571,577 ****
      }
  }
  \f
! #ifdef GET_LONGJMP_TARGET
  /* Figure out where the longjmp will land.  Slurp the args out of the stack.
     We expect the first arg to be a pointer to the jmp_buf structure from which
     we extract the pc (JB_PC) that we will land at.  The pc is copied into PC.
--- 571,579 ----
      }
  }
  \f
! /* Because of Multi-arch, GET_LONGJMP_TARGET is always defined.  So test
!    for a definition of JB_PC.  */
! #ifdef JB_PC
  /* Figure out where the longjmp will land.  Slurp the args out of the stack.
     We expect the first arg to be a pointer to the jmp_buf structure from which
     we extract the pc (JB_PC) that we will land at.  The pc is copied into PC.
Index: sparc-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/sparc-tdep.c,v
retrieving revision 1.22
diff -p -r1.22 sparc-tdep.c
*** sparc-tdep.c	2002/01/20 19:26:48	1.22
--- sparc-tdep.c	2002/02/19 10:35:26
*************** fill_fpregset (gdb_fpregset_t *fpregsetp
*** 1686,1693 ****
  
  #endif /* USE_PROC_FS */
  
! 
! #ifdef GET_LONGJMP_TARGET
  
  /* Figure out where the longjmp will land.  We expect that we have just entered
     longjmp and haven't yet setup the stack frame, so the args are still in the
--- 1686,1694 ----
  
  #endif /* USE_PROC_FS */
  
! /* Because of Multi-arch, GET_LONGJMP_TARGET is always defined.  So test
!    for a definition of JB_PC.  */
! #ifdef JB_PC
  
  /* Figure out where the longjmp will land.  We expect that we have just entered
     longjmp and haven't yet setup the stack frame, so the args are still in the