Patchwork GDB/opcodes: Remove arch/mach/endian disassembler assertions

login
register
mail settings
Submitter Maciej W. Rozycki
Date Aug. 1, 2017, 4:39 p.m.
Message ID <alpine.DEB.2.00.1708010044560.29991@tp.orcam.me.uk>
Download mbox | patch
Permalink /patch/21841/
State Committed
Headers show

Comments

Maciej W. Rozycki - Aug. 1, 2017, 4:39 p.m.
Fix `set architecture' and `set endian' command disassembly regressions 
from commit 39503f82427e ("Delegate opcodes to select disassembler in 
GDB"), and commit 003ca0fd2286 ("Refactor disassembler selection"), as 
well as a MIPS compressed ISA disassembly target regression from commit 
6394c606997f ("Don't use print_insn_XXX in GDB"), which caused assertion 
failures to trigger.

For example with the `mips-linux-gnu' target we get:

$ cat main.c
int
main (void)
{
  return 0;
}
$ gcc -mips32r2 -O2 main.c -o main
$ gcc -mips16 -mips32r2 -O2 main.c -o main16
$ gdb
GNU gdb (GDB) 8.0.50.20170731-git
[...]
(gdb) file main
Reading symbols from main...done.
(gdb) show architecture
The target architecture is set automatically (currently mips:isa32r2)
(gdb) show endian
The target endianness is set automatically (currently big endian)
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400500 <+0>:	jr	ra
   0x00400504 <+4>:	move	v0,zero
End of assembler dump.
(gdb) set architecture mips:isa64r2
The target architecture is assumed to be mips:isa64r2
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400500 <+0>:	
.../gdb/arch-utils.c:979: internal-error: int default_print_insn(bfd_vma, disassemble_info*): Assertion `info->mach == bfd_get_mach (exec_bfd)' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) n
[...]
Command aborted.
(gdb) set architecture auto
The target architecture is set automatically (currently mips:isa32r2)
(gdb) set endian little
The target is assumed to be little endian
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400500 <+0>:
.../gdb/arch-utils.c:978: internal-error: int default_print_insn(bfd_vma, disassemble_info*): Assertion `info->endian == (bfd_big_endian (exec_bfd) ? BFD_ENDIAN_BIG : BFD_ENDIAN_LITTLE)' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) n
[...]
Command aborted.
(gdb) set endian auto
The target endianness is set automatically (currently big endian)
(gdb) set architecture i386
The target architecture is assumed to be i386
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400500 <+0>:
.../gdb/arch-utils.c:976: internal-error: int default_print_insn(bfd_vma, disassemble_info*): Assertion `info->arch == bfd_get_arch (exec_bfd)' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) n
[...]
Command aborted.
(gdb) set architecture auto
The target architecture is set automatically (currently mips:isa32r2)
(gdb) file main16
Load new symbol table from "main16"? (y or n) y
Reading symbols from main16...done.
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400501 <+0>:	
.../gdb/arch-utils.c:979: internal-error: int default_print_insn(bfd_vma, disassemble_info*): Assertion `info->mach == bfd_get_mach (exec_bfd)' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) n
Command aborted.
(gdb) 

Remove the assertions then, restoring previous semantics:

(gdb) file main
Reading symbols from main...done.
(gdb) set architecture mips:isa64r2
The target architecture is assumed to be mips:isa64r2
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400500 <+0>:	jr	ra
   0x00400504 <+4>:	move	v0,zero
End of assembler dump.
(gdb) set endian little
The target is assumed to be little endian
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400500 <+0>:	j	0x3800c
   0x00400504 <+4>:	addiu	s0,t0,0
End of assembler dump.
(gdb) set architecture i386
The target architecture is assumed to be i386
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400500 <+0>:	add    %eax,%esp
   0x00400502 <+2>:	add    %cl,(%eax)
   0x00400504 <+4>:	add    %al,(%eax)
   0x00400506 <+6>:	adc    %ah,0x0
End of assembler dump.
(gdb) set architecture auto
The target architecture is set automatically (currently mips:isa32r2)
(gdb) set endian auto
The target endianness is set automatically (currently big endian)
(gdb) file main16
Load new symbol table from "main16"? (y or n) y
Reading symbols from main16...done.
(gdb) disassemble main
Dump of assembler code for function main:
   0x00400501 <+0>:	jr	ra
   0x00400503 <+2>:	li	v0,0
End of assembler dump.
(gdb) 

	gdb/
	* arch-utils.c (default_print_insn): Remove arch/mach/endian
	assertions.

	opcodes/
	* disassemble.c (disassembler): Remove arch/mach/endian 
	assertions.
---
Hi,

 Regression-tested with GDB using the `mips-linux-gnu' target, o32 ABI, 
bringing the number of failures down by ~300.  Also no regressions between 
plain commit 6394c606997f^ and commit 6394c606997f with this patch applied 
(as long as the mem-break fix, posted separately, has been also applied).

 OK to apply.

  Maciej

---
 gdb/arch-utils.c      |    7 -------
 opcodes/disassemble.c |   13 +------------
 2 files changed, 1 insertion(+), 19 deletions(-)

gdb-opcode-disassemble-assert.diff
Alan Modra - Aug. 7, 2017, 1:24 p.m.
On Tue, Aug 01, 2017 at 05:39:33PM +0100, Maciej W. Rozycki wrote:
> 	gdb/
> 	* arch-utils.c (default_print_insn): Remove arch/mach/endian
> 	assertions.
> 
> 	opcodes/
> 	* disassemble.c (disassembler): Remove arch/mach/endian 
> 	assertions.

This is OK, obvious really.
Maciej W. Rozycki - Aug. 7, 2017, 2:56 p.m.
On Mon, 7 Aug 2017, Alan Modra wrote:

> > 	gdb/
> > 	* arch-utils.c (default_print_insn): Remove arch/mach/endian
> > 	assertions.
> > 
> > 	opcodes/
> > 	* disassemble.c (disassembler): Remove arch/mach/endian 
> > 	assertions.
> 
> This is OK, obvious really.

 Applied now, thanks for your review.

  Maciej

Patch

Index: binutils/gdb/arch-utils.c
===================================================================
--- binutils.orig/gdb/arch-utils.c	2017-07-31 16:21:27.410997829 +0100
+++ binutils/gdb/arch-utils.c	2017-07-31 16:22:15.571550657 +0100
@@ -971,13 +971,6 @@  default_print_insn (bfd_vma memaddr, dis
 {
   disassembler_ftype disassemble_fn;
 
-  if (exec_bfd != NULL)
-    {
-      gdb_assert (info->arch == bfd_get_arch (exec_bfd));
-      gdb_assert (info->endian == (bfd_big_endian (exec_bfd)
-				   ? BFD_ENDIAN_BIG : BFD_ENDIAN_LITTLE));
-      gdb_assert (info->mach == bfd_get_mach (exec_bfd));
-    }
   disassemble_fn = disassembler (info->arch, info->endian == BFD_ENDIAN_BIG,
 				 info->mach, exec_bfd);
 
Index: binutils/opcodes/disassemble.c
===================================================================
--- binutils.orig/opcodes/disassemble.c	2017-07-31 11:12:12.000000000 +0100
+++ binutils/opcodes/disassemble.c	2017-07-31 16:23:47.294937646 +0100
@@ -111,21 +111,10 @@ 
 
 disassembler_ftype
 disassembler (enum bfd_architecture a, bfd_boolean big, unsigned long mach,
-	      bfd *abfd)
+	      bfd *abfd ATTRIBUTE_UNUSED)
 {
   disassembler_ftype disassemble;
 
-  if (abfd != NULL)
-    {
-      /* Do some asserts that the first three parameters should equal
-	 to what we can get from ABFD.  On the other hand, these
-	 asserts help removing some compiler errors on unused
-	 parameter.  */
-      assert (a == bfd_get_arch (abfd));
-      assert (big == bfd_big_endian (abfd));
-      assert (mach == bfd_get_mach (abfd));
-    }
-
   switch (a)
     {
       /* If you add a case to this table, also add it to the