Fix gdb.multi/base.exp failures.

UNRESOLVED: gdb.multi/base.exp: remove-inferiors 2-3
UNRESOLVED: gdb.multi/base.exp: check remove-inferiors

gdb is crashing because it's accessing/freeing already freed memory.

==16368== Invalid read of size 4
==16368==    at 0x660A9D: find_pc_section (binutils-gdb/gdb/objfiles.c:1349)
==16368==    by 0x663ECB: lookup_minimal_symbol_by_pc_section (binutils-gdb/gdb/minsyms.c:734)
==16368==    by 0x5D987A: find_pc_sect_symtab (binutils-gdb/gdb/symtab.c:2153)
==16368==    by 0x5D4D77: blockvector_for_pc_sect (binutils-gdb/gdb/block.c:168)
==16368==    by 0x5D4F59: block_for_pc_sect (binutils-gdb/gdb/block.c:246)
==16368==    by 0x5D4F9B: block_for_pc (binutils-gdb/gdb/block.c:258)
==16368==    by 0x734C5D: inline_frame_sniffer (binutils-gdb/gdb/inline-frame.c:218)
==16368==    by 0x732104: frame_unwind_try_unwinder (binutils-gdb/gdb/frame-unwind.c:108)
==16368==    by 0x73223F: frame_unwind_find_by_frame (binutils-gdb/gdb/frame-unwind.c:159)
==16368==    by 0x72D5AA: compute_frame_id (binutils-gdb/gdb/frame.c:453)
==16368==    by 0x7300EC: get_prev_frame_if_no_cycle (binutils-gdb/gdb/frame.c:1758)
==16368==    by 0x73079A: get_prev_frame_always (binutils-gdb/gdb/frame.c:1931)
==16368==  Address 0x5b13500 is 16 bytes inside a block of size 24 free'd
==16368==    at 0x403072E: free (valgrind/coregrind/m_replacemalloc/vg_replace_malloc.c:445)
==16368==    by 0x762134: xfree (binutils-gdb/gdb/common/common-utils.c:108)
==16368==    by 0x65DACF: objfiles_pspace_data_cleanup (binutils-gdb/gdb/objfiles.c:91)
==16368==    by 0x75E546: program_spaceregistry_callback_adaptor (binutils-gdb/gdb/progspace.c:45)
==16368==    by 0x7644F6: registry_clear_data (binutils-gdb/gdb/registry.c:82)
==16368==    by 0x7645AB: registry_container_free_data (binutils-gdb/gdb/registry.c:95)
==16368==    by 0x75E5B4: program_space_free_data (binutils-gdb/gdb/progspace.c:45)
==16368==    by 0x75E9BA: release_program_space (binutils-gdb/gdb/progspace.c:167)
==16368==    by 0x75EB9B: prune_program_spaces (binutils-gdb/gdb/progspace.c:269)
==16368==    by 0x75303D: remove_inferior_command (binutils-gdb/gdb/inferior.c:792)
==16368==    by 0x50B5FD: do_cfunc (binutils-gdb/gdb/cli/cli-decode.c:107)
==16368==    by 0x50E6F2: cmd_func (binutils-gdb/gdb/cli/cli-decode.c:1886)

The problem originates from the get_current_arch call in
py-progspace.c:py_free_pspace.  The inferior associated with the
pspace is gone, and the current inferior is a different one and is running.
Therefore get_current_arch tries to read the current frame which
causes reads of data in the current program space which we've just deleted.

	* python/py-progspace.c (py_free_pspace): Call target_gdbarch
	instead of get_current_arch.

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index ece66c261e7649ac3e0806051e65b5b3f68e3036..0c87bfff3c5981284d845354ee88cae8fb1866e1 100644 (file)
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,8 @@
+2014-05-20  Doug Evans  <dje@google.com>
+
+       * python/py-progspace.c (py_free_pspace): Call target_gdbarch
+       instead of get_current_arch.
+
 2014-05-20  Pedro Alves  <palves@redhat.com>
 
        * NEWS: Mention that compare-sections now works with all targets.

diff --git a/gdb/python/py-progspace.c b/gdb/python/py-progspace.c
index cda5a86ff91b1388b2a634fb7dc919cf7f3da747..db4c564ac222cd2175a7451d3fd74436de889e2c 100644 (file)
--- a/gdb/python/py-progspace.c
+++ b/gdb/python/py-progspace.c
@@ -241,7 +241,16 @@ py_free_pspace (struct program_space *pspace, void *datum)
 {
   struct cleanup *cleanup;
   pspace_object *object = datum;
-  struct gdbarch *arch = get_current_arch ();
+  /* This is a fiction, but we're in a nasty spot: The pspace is in the
+     process of being deleted, we can't rely on anything in it.  Plus
+     this is one time when the current program space and current inferior
+     are not in sync: All inferiors that use PSPACE may no longer exist.
+     We don't need to do much here, and since "there is always an inferior"
+     using target_gdbarch suffices.
+     Note: We cannot call get_current_arch because it may try to access
+     the target, which may involve accessing data in the pspace currently
+     being deleted.  */
+  struct gdbarch *arch = target_gdbarch ();
 
   cleanup = ensure_python_env (arch, current_language);
   object->pspace = NULL;