Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Mark Kettenis <kettenis@wins.uva.nl>
To: jimb@zwingli.cygnus.com
Cc: gdb-patches@sources.redhat.com
Subject: Re: [PING][RFA] Fall back on dynamic symtab in solib.c:bdf_lookup_symbol
Date: Mon, 07 Aug 2000 03:59:00 -0000	[thread overview]
Message-ID: <200008071059.e77AxPb00589@delius.kettenis.local> (raw)
In-Reply-To: <npog3ikmuy.fsf@zwingli.cygnus.com>

   From: Jim Blandy <jimb@zwingli.cygnus.com>
   Date: 28 Jul 2000 15:42:45 -0500

   > I'm still looking for approval of the following patch:
   > 
   >   http://sources.redhat.com/ml/gdb-patches/2000-05/msg00318.html
   > 
   > Mark

   Sorry about the delay.  If this causes no test suite failures, then
   it's fine to commit it.

Checked in now.

Mark
From taylor@cygnus.com Mon Aug 07 07:26:00 2000
From: David Taylor <taylor@cygnus.com>
To: gdb-patches@sourceware.cygnus.com
Subject: [PATCH] bug in build_parse (parse.c)
Date: Mon, 07 Aug 2000 07:26:00 -0000
Message-id: <200008071425.KAA05946@texas.cygnus.com>
X-SW-Source: 2000-08/msg00138.html
Content-length: 2968

In build_parse, you'll find the lines:

      num_std_regs = 0;
    #ifdef PC_REGNUM
      if (PC_REGNUM >= 0)
	num_std_regs++;
    #endif
    #ifdef FP_REGNUM
      if (FP_REGNUM >= 0)
	num_std_regs++;
    #endif
    #ifdef SP_REGNUM
      if (SP_REGNUM >= 0)
	num_std_regs++;
    #endif
    #ifdef PS_REGNUM
      if (PS_REGNUM >= 0)
	num_std_regs++;
    #endif

Notice how num_std_regs only gets incremented if {PC_REGNUM,
FP_REGNUM, SP_REGNUM, PS_REGNUM} is *BOTH* defined and >= 0.

      /* create an empty table */
      std_regs = xmalloc ((num_std_regs + 1) * sizeof *std_regs);

But notice how it gets filled in if it's defined, regardless of
whether it is >= 0 or not:

      i = 0;
      /* fill it in */
    #ifdef PC_REGNUM
      std_regs[i].name = "pc";
      std_regs[i].regnum = PC_REGNUM;
      i++;
    #endif
    #ifdef FP_REGNUM
      std_regs[i].name = "fp";
      std_regs[i].regnum = FP_REGNUM;
      i++;
    #endif
    #ifdef SP_REGNUM
      std_regs[i].name = "sp";
      std_regs[i].regnum = SP_REGNUM;
      i++;
    #endif
    #ifdef PS_REGNUM
      std_regs[i].name = "ps";
      std_regs[i].regnum = PS_REGNUM;
      i++;
    #endif

Thus, if one of more of them is defined (possibly because other parts
of GDB won't compile if it's not defined), but is defined as -1
(because the processor doesn't support it of there is no ABI or...),
then build_parse will write off the end of the std_regs array.

Ooops.  Here's a patch to fix it:

Index: parse.c
===================================================================
RCS file: /cvs/src/src/gdb/parse.c,v
retrieving revision 1.8
diff -c -r1.8 parse.c
*** parse.c     2000/07/30 01:48:26     1.8
--- parse.c     2000/08/07 14:20:31
***************
*** 1331,1354 ****
    i = 0;
    /* fill it in */
  #ifdef PC_REGNUM
!   std_regs[i].name = "pc";
!   std_regs[i].regnum = PC_REGNUM;
!   i++;
  #endif
  #ifdef FP_REGNUM
!   std_regs[i].name = "fp";
!   std_regs[i].regnum = FP_REGNUM;
!   i++;
  #endif
  #ifdef SP_REGNUM
!   std_regs[i].name = "sp";
!   std_regs[i].regnum = SP_REGNUM;
!   i++;
  #endif
  #ifdef PS_REGNUM
!   std_regs[i].name = "ps";
!   std_regs[i].regnum = PS_REGNUM;
!   i++;
  #endif
    memset (&std_regs[i], 0, sizeof (std_regs[i]));
  }
--- 1331,1366 ----
    i = 0;
    /* fill it in */
  #ifdef PC_REGNUM
!   if (PC_REGNUM >= 0)
!     {
!       std_regs[i].name = "pc";
!       std_regs[i].regnum = PC_REGNUM;
!       i++;
!     }
  #endif
  #ifdef FP_REGNUM
!   if (FP_REGNUM >= 0)
!     {
!       std_regs[i].name = "fp";
!       std_regs[i].regnum = FP_REGNUM;
!       i++;
!     }
  #endif
  #ifdef SP_REGNUM
!   if (SP_REGNUM >= 0)
!     {
!       std_regs[i].name = "sp";
!       std_regs[i].regnum = SP_REGNUM;
!       i++;
!     }
  #endif
  #ifdef PS_REGNUM
!   if (PS_REGNUM >= 0)
!     {
!       std_regs[i].name = "ps";
!       std_regs[i].regnum = PS_REGNUM;
!       i++;
!     }
  #endif
    memset (&std_regs[i], 0, sizeof (std_regs[i]));
  }
From ezannoni@cygnus.com Mon Aug 07 08:03:00 2000
From: Elena Zannoni <ezannoni@cygnus.com>
To: gdb-patches@sources.redhat.com
Subject: [PATCH] Fix problem with missing .bss
Date: Mon, 07 Aug 2000 08:03:00 -0000
Message-id: <14734.53102.598696.678240@kwikemart.cygnus.com>
X-SW-Source: 2000-08/msg00139.html
Content-length: 8197

On Cygwin, sometimes the objfiles may not have a .bss section.  In
this case, the sect_index_bss would be not valid (== -1) and the
SECT_OFF_BSS macro would trigger an internal error.

This patch fixes that problem. I am committing it now.

Elena

2000-08-07  Elena Zannoni  <ezannoni@kwikemart.cygnus.com>

	* objfiles.h (SECT_OFF_BSS): Don't detect invalid sect_index_bss
 	here, let the users of the macro do it.
	* symtab.h (ANOFFSET): Detect here if the section index is not
 	initialized.
	* xcoffread.c (find_targ_sec): Don't treat .bss as special,
 	because some objfiles may not have that section at all.
	* coffread.c (cs_to_section): Ditto.
	* elfread.c (elf_symtab_read): Detect an uninitialized index
 	value.
	(elfstab_offset_sections): The macro ANOFFSET cannot be used as an
 	lvalue anymore.
	* remote.c (get_offsets, remote_cisco_objfile_relocate): Don't use
 	ANOFFSET as an lvalue.
	* objfiles.c (objfile_relocate, objfile_relocate): Don't use
 	ANOFFSET as an lvalue.
	* symfile.c (default_symfile_offsets): Don't use ANOFFSET as an
 	lvalue.

Index: coffread.c
===================================================================
RCS file: /cvs/src/src/gdb/coffread.c,v
retrieving revision 1.10
diff -c -u -p -r1.10 coffread.c
--- coffread.c	2000/07/30 01:48:24	1.10
+++ coffread.c	2000/08/07 14:47:34
@@ -347,7 +347,8 @@ cs_to_section (struct coff_symbol *cs, s
       else if (bfd_get_section_flags (abfd, sect) & SEC_LOAD)
 	off = SECT_OFF_DATA (objfile);
       else
-	off = SECT_OFF_BSS (objfile);
+	/* Just return the bfd section index. */
+	off = sect->index;
     }
   return off;
 }
Index: elfread.c
===================================================================
RCS file: /cvs/src/src/gdb/elfread.c,v
retrieving revision 1.10
diff -c -u -p -r1.10 elfread.c
--- elfread.c	2000/07/30 01:48:25	1.10
+++ elfread.c	2000/08/07 14:47:34
@@ -474,11 +474,16 @@ elf_symtab_read (struct objfile *objfile
 				    (char *) filesym->name;
 				}
 			    }
-			  if (sectinfo->sections[index] != 0)
-			    {
-			      complain (&section_info_dup_complaint,
-					sectinfo->filename);
+			  if (index != -1)
+			    { 
+			      if (sectinfo->sections[index] != 0)
+				{
+				  complain (&section_info_dup_complaint,
+					    sectinfo->filename);
+				}
 			    }
+			  else
+			    internal_error ("Section index uninitialized.");
 			  /* Bfd symbols are section relative. */
 			  symaddr = sym->value + sym->section->vma;
 			  /* Relocate non-absolute symbols by the section offset. */
@@ -486,7 +491,10 @@ elf_symtab_read (struct objfile *objfile
 			    {
 			      symaddr += offset;
 			    }
-			  sectinfo->sections[index] = symaddr;
+			  if (index != -1)
+			    sectinfo->sections[index] = symaddr;
+			  else
+			    internal_error ("Section index uninitialized.");
 			  /* The special local symbols don't go in the
 			     minimal symbol table, so ignore this one. */
 			  continue;
@@ -790,7 +798,7 @@ elfstab_offset_sections (struct objfile 
       pst->section_offsets = (struct section_offsets *)
 	obstack_alloc (&objfile->psymbol_obstack, SIZEOF_SECTION_OFFSETS);
       for (i = 0; i < SECT_OFF_MAX; i++)
-	ANOFFSET (pst->section_offsets, i) = maybe->sections[i];
+	(pst->section_offsets)->offsets[i] = maybe->sections[i];
       return;
     }
 
Index: objfiles.c
===================================================================
RCS file: /cvs/src/src/gdb/objfiles.c,v
retrieving revision 1.8
diff -c -u -p -r1.8 objfiles.c
--- objfiles.c	2000/08/04 23:13:49	1.8
+++ objfiles.c	2000/08/07 14:47:35
@@ -520,7 +520,7 @@ objfile_relocate (struct objfile *objfil
     int something_changed = 0;
     for (i = 0; i < objfile->num_sections; ++i)
       {
-	ANOFFSET (delta, i) =
+	delta->offsets[i] =
 	  ANOFFSET (new_offsets, i) - ANOFFSET (objfile->section_offsets, i);
 	if (ANOFFSET (delta, i) != 0)
 	  something_changed = 1;
@@ -639,7 +639,7 @@ objfile_relocate (struct objfile *objfil
   {
     int i;
     for (i = 0; i < objfile->num_sections; ++i)
-      ANOFFSET (objfile->section_offsets, i) = ANOFFSET (new_offsets, i);
+      (objfile->section_offsets)->offsets[i] = ANOFFSET (new_offsets, i);
   }
 
   {
Index: objfiles.h
===================================================================
RCS file: /cvs/src/src/gdb/objfiles.h,v
retrieving revision 1.5
diff -c -u -p -r1.5 objfiles.h
--- objfiles.h	2000/05/28 01:12:28	1.5
+++ objfiles.h	2000/08/07 14:47:35
@@ -597,8 +597,9 @@ extern int is_in_import_list (char *, st
      ((objfile->sect_index_text == -1) ? \
       (internal_error ("sect_index_text not initialized"), -1) : objfile->sect_index_text)
 
-#define SECT_OFF_BSS(objfile) \
-     ((objfile->sect_index_bss == -1) ? \
-      (internal_error ("sect_index_bss not initialized"), -1) : objfile->sect_index_bss)
+/* Sometimes the .bss section is missing from the objfile, so we don't
+   want to die here. Let the users of SECT_OFF_BSS deal with an
+   uninitialized section index. */
+#define SECT_OFF_BSS(objfile) (objfile)->sect_index_bss
 
 #endif /* !defined (OBJFILES_H) */
Index: remote.c
===================================================================
RCS file: /cvs/src/src/gdb/remote.c,v
retrieving revision 1.18
diff -c -u -p -r1.18 remote.c
--- remote.c	2000/08/04 09:32:19	1.18
+++ remote.c	2000/08/07 14:47:36
@@ -1835,14 +1835,14 @@ get_offsets (void)
   offs = (struct section_offsets *) alloca (SIZEOF_SECTION_OFFSETS);
   memcpy (offs, symfile_objfile->section_offsets, SIZEOF_SECTION_OFFSETS);
 
-  ANOFFSET (offs, SECT_OFF_TEXT (symfile_objfile)) = text_addr;
+  offs->offsets[SECT_OFF_TEXT (symfile_objfile)] = text_addr;
 
   /* This is a temporary kludge to force data and bss to use the same offsets
      because that's what nlmconv does now.  The real solution requires changes
      to the stub and remote.c that I don't have time to do right now.  */
 
-  ANOFFSET (offs, SECT_OFF_DATA (symfile_objfile)) = data_addr;
-  ANOFFSET (offs, SECT_OFF_BSS (symfile_objfile)) = data_addr;
+  offs->offsets[SECT_OFF_DATA (symfile_objfile)] = data_addr;
+  offs->offsets[SECT_OFF_BSS (symfile_objfile)] = data_addr;
 
   objfile_relocate (symfile_objfile, offs);
 }
@@ -1948,9 +1948,9 @@ remote_cisco_objfile_relocate (bfd_signe
       offs = (struct section_offsets *) alloca (SIZEOF_SECTION_OFFSETS);
       memcpy (offs, symfile_objfile->section_offsets, SIZEOF_SECTION_OFFSETS);
 
-      ANOFFSET (offs, SECT_OFF_TEXT (symfile_objfile)) = text_off;
-      ANOFFSET (offs, SECT_OFF_DATA (symfile_objfile)) = data_off;
-      ANOFFSET (offs, SECT_OFF_BSS (symfile_objfile)) = bss_off;
+      offs->offsets[SECT_OFF_TEXT (symfile_objfile)] = text_off;
+      offs->offsets[SECT_OFF_DATA (symfile_objfile)] = data_off;
+      offs->offsets[SECT_OFF_BSS (symfile_objfile)] = bss_off;
 
       /* First call the standard objfile_relocate.  */
       objfile_relocate (symfile_objfile, offs);
Index: symfile.c
===================================================================
RCS file: /cvs/src/src/gdb/symfile.c,v
retrieving revision 1.16
diff -c -u -p -r1.16 symfile.c
--- symfile.c	2000/07/30 01:48:27	1.16
+++ symfile.c	2000/08/07 14:47:37
@@ -521,7 +521,7 @@ default_symfile_offsets (struct objfile 
       /* Record all sections in offsets */
       /* The section_offsets in the objfile are here filled in using
          the BFD index. */
-      ANOFFSET (objfile->section_offsets, osp->sectindex) = osp->addr;
+      (objfile->section_offsets)->offsets[osp->sectindex] = osp->addr;
     }
 
   /* Remember the bfd indexes for the .text, .data, .bss and
Index: symtab.h
===================================================================
RCS file: /cvs/src/src/gdb/symtab.h,v
retrieving revision 1.12
diff -c -u -p -r1.12 symtab.h
--- symtab.h	2000/08/04 23:13:50	1.12
+++ symtab.h	2000/08/07 14:47:38
@@ -829,7 +829,9 @@ struct section_offsets
     CORE_ADDR offsets[1];	/* As many as needed. */
   };
 
-#define	ANOFFSET(secoff, whichone)	(secoff->offsets[whichone])
+#define	ANOFFSET(secoff, whichone) \
+   ((whichone == -1) ? \
+    (internal_error ("Section index is uninitialized"), -1) : secoff->offsets[whichone])
 
 /* The maximum possible size of a section_offsets table.  */
 


      reply	other threads:[~2000-08-07  3:59 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <200007270710.e6R7AjS00711@delius.kettenis.local>
2000-08-04 14:43 ` Jim Blandy
2000-08-07  3:59   ` Mark Kettenis [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=200008071059.e77AxPb00589@delius.kettenis.local \
    --to=kettenis@wins.uva.nl \
    --cc=gdb-patches@sources.redhat.com \
    --cc=jimb@zwingli.cygnus.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox