[patchv4 3/5] new: solib-sunos.c objfile unification [test needed]

[patchv4 3/5] new: solib-sunos.c objfile unification [test needed]

[Jan Kratochvil]

Hi,

I would appreciate testing this (probably a full testsuite diff needed) on
Solaris.  I found some local older Opensolaris VM here but recent GDB fails to
build there (on gnulib fnmatch*); according to Joel recent GDBs build fine on
real Solaris.

The reason is apparently that objfile should be created at a single place,
otherwise various changes need to be done at multiple places etc.

But unfortunately I do not see the reason why the code did not use
allocate_objfile() already.  It was added this way by this release from
https://github.com/palves/gdb-old-releases.git but I do not know why:
	commit e62555e9d0cf7be157ac62cc31a3d1d079ba271b
	Date:   Wed Feb 1 19:21:24 1995 +0100
	gdb-4.14+gdb-4.14-testsuite

also in archer.git jankratochvil/dwp


Thanks,
Jan


gdb/
2013-09-18  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* solib-sunos.c (allocate_rt_common_objfile): Replace it by an
	allocate_objfile call.
	(solib_add_common_symbols): Call free_objfile and allocate_objfile
	instead.

--- a/gdb/solib-sunos.c
+++ b/gdb/solib-sunos.c
@@ -182,31 +182,11 @@ static int match_main (char *);
 static void
 allocate_rt_common_objfile (void)
 {
-  struct objfile *objfile;
-  struct objfile *last_one;
-
-  objfile = (struct objfile *) xmalloc (sizeof (struct objfile));
-  memset (objfile, 0, sizeof (struct objfile));
-  objfile->psymbol_cache = psymbol_bcache_init ();
-  objfile->macro_cache = bcache_xmalloc (NULL, NULL);
-  objfile->filename_cache = bcache_xmalloc (NULL, NULL);
-  obstack_init (&objfile->objfile_obstack);
-  objfile->name = xstrdup ("rt_common");
-
-  /* Add this file onto the tail of the linked list of other such files.  */
-
-  objfile->next = NULL;
-  if (object_files == NULL)
-    object_files = objfile;
-  else
-    {
-      for (last_one = object_files;
-	   last_one->next;
-	   last_one = last_one->next);
-      last_one->next = objfile;
-    }
+  const char name[] = "rt_common";
 
-  rt_common_objfile = objfile;
+  rt_common_objfile = allocate_objfile (NULL, 0);
+  rt_common_objfile->name = obstack_copy0 (&rt_common_objfile->objfile_obstack,
+					   name, strlen (name));
 }
 
 /* Read all dynamically loaded common symbol definitions from the inferior
@@ -225,10 +205,8 @@ solib_add_common_symbols (CORE_ADDR rtc_symp)
 
   if (rt_common_objfile != NULL && rt_common_objfile->minimal_symbol_count)
     {
-      obstack_free (&rt_common_objfile->objfile_obstack, 0);
-      obstack_init (&rt_common_objfile->objfile_obstack);
-      rt_common_objfile->minimal_symbol_count = 0;
-      rt_common_objfile->msymbols = NULL;
+      free_objfile (rt_common_objfile);
+      allocate_rt_common_objfile ();
       terminate_minimal_symbol_table (rt_common_objfile);
     }
 

Re: [patchv4 3/5] new: solib-sunos.c objfile unification [test needed]

[Joel Brobecker]

> gdb/
> 2013-09-18  Jan Kratochvil  <jan.kratochvil@redhat.com>
> 
> 	* solib-sunos.c (allocate_rt_common_objfile): Replace it by an
> 	allocate_objfile call.
> 	(solib_add_common_symbols): Call free_objfile and allocate_objfile
> 	instead.

If this is really used on Solaris, I wouldn't mind testing it
(though only with our internal testing, as something in the official
testing badly crashes every single of our Solaris machine).
But isn't this file only used in the case of (old) SunOS, that is
pre-Solaris? Actually, a quick grep shows:

./config/vax/nbsdaout.mh:       solib-sunos.o
./config/sparc/nbsdaout.mh:     sparc-nat.o sparcnbsd-nat.o bsd-kvm.o solib-sunos.o
./config/m68k/nbsdaout.mh:      solib-sunos.o
./config/m68k/obsd.mh:  solib-sunos.o
./config/arm/nbsdaout.mh:       solib-sunos.o
./config/i386/nbsdaout.mh:      solib-sunos.o
./config/i386/obsdaout.mh:      solib-sunos.o

I must be missing something, or else did you build with --enable-targets=all?

-- 
Joel

Re: [patchv4 3/5] new: solib-sunos.c objfile unification [test needed]

[Jan Kratochvil]

On Thu, 19 Sep 2013 16:19:31 +0200, Joel Brobecker wrote:
> But isn't this file only used in the case of (old) SunOS, that is
> pre-Solaris? Actually, a quick grep shows:
> 
> ./config/vax/nbsdaout.mh:       solib-sunos.o
> ./config/sparc/nbsdaout.mh:     sparc-nat.o sparcnbsd-nat.o bsd-kvm.o solib-sunos.o
> ./config/m68k/nbsdaout.mh:      solib-sunos.o
> ./config/m68k/obsd.mh:  solib-sunos.o
> ./config/arm/nbsdaout.mh:       solib-sunos.o
> ./config/i386/nbsdaout.mh:      solib-sunos.o
> ./config/i386/obsdaout.mh:      solib-sunos.o
> 
> I must be missing something, or else did you build with --enable-targets=all?

OK, I did not expect that.  Yes, I always build with --enable-targets=all.

Is anyone using this target?  I think the patch is OK to check-in I think.


Thanks,
Jan

Re: [patchv4 3/5] new: solib-sunos.c objfile unification [test needed]

[Joel Brobecker]

> OK, I did not expect that.  Yes, I always build with --enable-targets=all.
> Is anyone using this target?  I think the patch is OK to check-in I think.

MarkK, maybe?

-- 
Joel

Re: [patchv4 3/5] new: solib-sunos.c objfile unification [test needed]

[Pedro Alves]

See:
  https://sourceware.org/ml/gdb/2013-08/msg00077.html
and:
  https://sourceware.org/ml/gdb/2013-08/msg00085.html

-- 
Pedro Alves

[cancel] [patchv4 3/5] new: solib-sunos.c objfile unification [test needed]

[Jan Kratochvil]

On Thu, 19 Sep 2013 16:41:06 +0200, Pedro Alves wrote:
> See:
>   https://sourceware.org/ml/gdb/2013-08/msg00077.html
> and:
>   https://sourceware.org/ml/gdb/2013-08/msg00085.html

Somehow forgot it already... thanks.

Therefore posted:
	[patch 1/2] Remove a.out NetBSD and OpenBSD targets
	https://sourceware.org/ml/gdb-patches/2013-09/msg00699.html
	[patch 2/2] Remove solib-sunos.c
	https://sourceware.org/ml/gdb-patches/2013-09/msg00700.html


Jan