[patch] More workarounds for gold/15021

[Doug Evans <dje at google dot com>]

Hi.
My previous workaround for gold/15021 was insufficient,
as the testcase shows.

Regression tested on amd64-linux with/without fission.

I will check this in in a few days if there are no objections.

2013-08-27  Doug Evans  <dje@google.com>

	* dwarf2read.c (queue_and_load_all_dwo_tus): New function.
	(queue_and_load_dwo_tu): New function.
	(lookup_dwo_signatured_type): Set per_cu.tu_read.
	(maybe_queue_comp_unit): Rename this_cu argument to dependent_cu.
	Make dependent_cu optional.
	(dw2_do_instantiate_symtab): If we just loaded a CU from a DWO,
	and an older .gdb_index is in use, queue and load all its TUs too.

	testsuite/
	* gdb.base/enumval.c (ZERO): New enum value.
	(main): Use it
	* gdb.base/enumval.exp: Test ability to print ZERO.

Index: dwarf2read.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2read.c,v
retrieving revision 1.830
diff -u -p -r1.830 dwarf2read.c
--- dwarf2read.c	27 Aug 2013 21:38:05 -0000	1.830
+++ dwarf2read.c	27 Aug 2013 21:39:10 -0000
@@ -1778,6 +1778,8 @@ static struct dwo_unit *lookup_dwo_comp_
 static struct dwo_unit *lookup_dwo_type_unit
   (struct signatured_type *, const char *, const char *);
 
+static void queue_and_load_all_dwo_tus (struct dwarf2_per_cu_data *);
+
 static void free_dwo_file_cleanup (void *);
 
 static void process_cu_includes (void);
@@ -2338,6 +2340,17 @@ dw2_do_instantiate_symtab (struct dwarf2
     {
       queue_comp_unit (per_cu, language_minimal);
       load_cu (per_cu);
+
+      /* If we just loaded a CU from a DWO, and we're working with an index
+	 that may badly handle TUs, load all the TUs in that DWO as well.
+	 http://sourceware.org/bugzilla/show_bug.cgi?id=15021  */
+      if (!per_cu->is_debug_types
+	  && per_cu->cu->dwo_unit != NULL
+	  && dwarf2_per_objfile->index_table != NULL
+	  && dwarf2_per_objfile->index_table->version <= 7
+	  /* DWP files aren't supported yet.  */
+	  && get_dwp_file () == NULL)
+	queue_and_load_all_dwo_tus (per_cu);
     }
 
   process_queue ();
@@ -4468,6 +4481,7 @@ lookup_dwo_signatured_type (struct dwarf
     return NULL;
 
   fill_in_sig_entry_from_dwo_entry (objfile, sig_entry, dwo_entry);
+  sig_entry->per_cu.tu_read = 1;
   return sig_entry;
 }
 
@@ -6940,8 +6954,9 @@ queue_comp_unit (struct dwarf2_per_cu_da
   dwarf2_queue_tail = item;
 }
 
-/* THIS_CU has a reference to PER_CU.  If necessary, load the new compilation
-   unit and add it to our queue.
+/* If PER_CU is not yet queued, add it to the queue.
+   If DEPENDENT_CU is non-NULL, it has a reference to PER_CU so add a
+   dependency.
    The result is non-zero if PER_CU was queued, otherwise the result is zero
    meaning either PER_CU is already queued or it is already loaded.
 
@@ -6949,7 +6964,7 @@ queue_comp_unit (struct dwarf2_per_cu_da
    The caller is required to load PER_CU if we return non-zero.  */
 
 static int
-maybe_queue_comp_unit (struct dwarf2_cu *this_cu,
+maybe_queue_comp_unit (struct dwarf2_cu *dependent_cu,
 		       struct dwarf2_per_cu_data *per_cu,
 		       enum language pretend_language)
 {
@@ -6965,7 +6980,8 @@ maybe_queue_comp_unit (struct dwarf2_cu 
 
   /* Mark the dependence relation so that we don't flush PER_CU
      too early.  */
-  dwarf2_add_dependence (this_cu, per_cu);
+  if (dependent_cu != NULL)
+    dwarf2_add_dependence (dependent_cu, per_cu);
 
   /* If it's already on the queue, we have nothing to do.  */
   if (per_cu->queued)
@@ -9828,6 +9844,55 @@ lookup_dwo_type_unit (struct signatured_
   return lookup_dwo_cutu (&this_tu->per_cu, dwo_name, comp_dir, this_tu->signature, 1);
 }
 
+/* Traversal function for queue_and_load_all_dwo_tus.  */
+
+static int
+queue_and_load_dwo_tu (void **slot, void *info)
+{
+  struct dwo_unit *dwo_unit = (struct dwo_unit *) *slot;
+  struct dwarf2_per_cu_data *per_cu = (struct dwarf2_per_cu_data *) info;
+  ULONGEST signature = dwo_unit->signature;
+  struct signatured_type *sig_type =
+    lookup_dwo_signatured_type (per_cu->cu, signature);
+
+  if (sig_type != NULL)
+    {
+      struct dwarf2_per_cu_data *sig_cu = &sig_type->per_cu;
+
+      /* We pass NULL for DEPENDENT_CU because we don't yet know if there's
+	 a real dependency of PER_CU on SIG_TYPE.  That is detected later
+	 while processing PER_CU.  */
+      if (maybe_queue_comp_unit (NULL, sig_cu, per_cu->cu->language))
+	load_full_type_unit (sig_cu);
+      VEC_safe_push (dwarf2_per_cu_ptr, per_cu->imported_symtabs, sig_cu);
+    }
+
+  return 1;
+}
+
+/* Queue all TUs contained in the DWO of PER_CU to be read in.
+   The DWO may have the only definition of the type, though it may not be
+   referenced anywhere in PER_CU.  Thus we have to load *all* its TUs.
+   http://sourceware.org/bugzilla/show_bug.cgi?id=15021  */
+
+static void
+queue_and_load_all_dwo_tus (struct dwarf2_per_cu_data *per_cu)
+{
+  struct dwo_unit *dwo_unit;
+  struct dwo_file *dwo_file;
+
+  gdb_assert (!per_cu->is_debug_types);
+  gdb_assert (get_dwp_file () == NULL);
+  gdb_assert (per_cu->cu != NULL);
+
+  dwo_unit = per_cu->cu->dwo_unit;
+  gdb_assert (dwo_unit != NULL);
+
+  dwo_file = dwo_unit->dwo_file;
+  if (dwo_file->tus != NULL)
+    htab_traverse_noresize (dwo_file->tus, queue_and_load_dwo_tu, per_cu);
+}
+
 /* Free all resources associated with DWO_FILE.
    Close the DWO file and munmap the sections.
    All memory should be on the objfile obstack.  */
Index: testsuite/gdb.base/enumval.c
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/enumval.c,v
retrieving revision 1.2
diff -u -p -r1.2 enumval.c
--- testsuite/gdb.base/enumval.c	1 Jan 2013 06:33:25 -0000	1.2
+++ testsuite/gdb.base/enumval.c	27 Aug 2013 21:39:10 -0000
@@ -17,6 +17,8 @@
 
 enum e { I, J = 0xffffffffU, K = 0xf000000000000000ULL } e = J, f = K;
 
+enum { ZERO };
+
 void
 dummy()
 {
@@ -26,5 +28,5 @@ int
 main(void)
 {
   dummy();
-  return 0;
+  return ZERO; /* This is here to ensure it survives into the debug info.  */
 }
Index: testsuite/gdb.base/enumval.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/enumval.exp,v
retrieving revision 1.4
diff -u -p -r1.4 enumval.exp
--- testsuite/gdb.base/enumval.exp	27 Jun 2013 18:50:30 -0000	1.4
+++ testsuite/gdb.base/enumval.exp	27 Aug 2013 21:39:10 -0000
@@ -70,3 +70,9 @@ gdb_test_multiple $test $test {
 	}
     }
 }
+
+# gold/15021
+# With -fdebug-types-section, Gold's .gdb_index entry for ZERO refers to the
+# CU, but the CU doesn't use the TU (type unit) that defines ZERO.
+# Thus gdb has to read in every TU for the CU.
+gdb_test "p ZERO" "ZERO"