[patch] Workaround gcc bug 49906

[Aleksandar Ristovski <aristovski@qnx.com>]

[-- Attachment #1: Type: text/plain, Size: 1490 bytes --]

Hello,

Apparently gcc has quite a bit of issues generating proper line 
information. One of particularly harmful ones are described in gcc bug 
49906. While on x86 they appear to be benign (save the odd line where 
breakpoint ends up being set), on arm it may cause gdb to not be able to 
step into such function (when software single stepping) or, worse even, 
breakpoint set on such function using "break <fnname>" syntax may be set 
on non reachable instruction thus never being hit.

The patch attached addresses this issue by performing further sanity 
checking and discarding results of skipping over prologue if there is 
any suspicion that the line information may be bogus/invalid/buggy.

In addition to the patch, I'm attaching new test. Full regression test 
shows no regressions (that is, the new test FAILs on current code, and 
passes with the patch).


Thanks,

Aleksandar Ristovski
QNX Software Systems


gdb/ChangeLog:


2011-10-25  Aleksandar Ristovski  <aristovski@qnx.com>

       * symtab.c (skip_prologue_sal): Introduced start_sal_orig to keep
       entry value.  Added check for line after the prologue.  If line
       information for the next line exists but appears to be before in
       the code, discard skip over prologue work and revert original sal.



gdb/testsuite/ChangeLog:

2011-10-25  Aleksandar Ristovski  <aristovski@qnx.com>

         * prologue-gccbug49906.exp: New test.
         * prologue.c (f_gccbuf49906): New function, call new function.


[-- Attachment #2: gccpr49906-workaround-lineinfo-HEAD-201110251328.patch --]
[-- Type: text/x-patch, Size: 2825 bytes --]

Index: gdb/symtab.c
===================================================================
RCS file: /cvs/src/src/gdb/symtab.c,v
retrieving revision 1.283
diff -u -p -r1.283 symtab.c
--- gdb/symtab.c	21 Jul 2011 15:13:29 -0000	1.283
+++ gdb/symtab.c	25 Oct 2011 17:28:46 -0000
@@ -2474,6 +2474,7 @@ skip_prologue_sal (struct symtab_and_lin
   struct gdbarch *gdbarch;
   struct block *b, *function_block;
   int force_skip, skip;
+  const struct symtab_and_line start_sal_orig = *sal;
 
   /* Do not change the SAL is PC was specified explicitly.  */
   if (sal->explicit_pc)
@@ -2627,6 +2628,29 @@ skip_prologue_sal (struct symtab_and_lin
       sal->line = SYMBOL_LINE (BLOCK_FUNCTION (function_block));
       sal->symtab = SYMBOL_SYMTAB (BLOCK_FUNCTION (function_block));
     }
+  else
+    {
+      /* Check if line number of apparent prologue end comes after
+	 the next line information.  If so, do not skip prologue
+	 as something odd has happened.  Probably gcc bug 49906.
+	 Instead of giving awkward location for the function start,
+	 give the previous good line (fuction start). */
+      const struct symtab_and_line prologue_sal_next
+	= find_pc_line (sal->end, 0);
+
+      if ((sym? (BLOCK_START (SYMBOL_BLOCK_VALUE (sym)) <= prologue_sal_next.end
+		&& prologue_sal_next.end < BLOCK_END (SYMBOL_BLOCK_VALUE (sym)))
+	  : (lookup_minimal_symbol_by_pc_section (prologue_sal_next.end,
+						  section)
+	     == lookup_minimal_symbol_by_pc_section (sal->pc, section)))
+	  && prologue_sal_next.line != 0
+	  && prologue_sal_next.symtab == sal->symtab
+	  && prologue_sal_next.line >= start_sal_orig.line
+	  && prologue_sal_next.line < sal->line)
+	{
+	  *sal = start_sal_orig;
+	}
+    }
 }
 
 /* If P is of the form "operator[ \t]+..." where `...' is
@@ -4432,7 +4456,24 @@ skip_prologue_using_sal (struct gdbarch 
 	  /* Assume that a consecutive SAL for the same (or larger)
 	     line mark the prologue -> body transition.  */
 	  if (sal.line >= prologue_sal.line)
-	    break;
+	    {
+	      /* If the sal following SAL has line number lower than
+		 SAL.LINE, then something is fishy.
+		 Possibly gcc bug 49906  */
+	      const struct symtab_and_line prologue_sal_next
+		= find_pc_line (sal.end, 0);
+
+	      if (prologue_sal_next.end < end_pc
+		  && prologue_sal_next.line != 0
+		  && prologue_sal_next.symtab == sal.symtab
+		  && prologue_sal_next.line >= prologue_sal.line
+		  && prologue_sal_next.line < sal.line)
+		  /* Fishy: we have something past prologue that appears
+		     in the source as before.  Be conservative and return
+		     start of the prologue.  That's the best guess anyway. */
+		prologue_sal.end = prologue_sal.pc;
+	      break;
+	    }
 
 	  /* The line number is smaller.  Check that it's from the
 	     same function, not something inlined.  If it's inlined,

[-- Attachment #3: gccpr49906-workaround-lineinfo-HEAD-test-201110251329.patch --]
[-- Type: text/x-patch, Size: 2817 bytes --]

Index: gdb/testsuite/gdb.base/prologue-gccbug49906.exp
===================================================================
RCS file: gdb/testsuite/gdb.base/prologue-gccbug49906.exp
diff -N gdb/testsuite/gdb.base/prologue-gccbug49906.exp
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ gdb/testsuite/gdb.base/prologue-gccbug49906.exp	25 Oct 2011 17:29:12 -0000
@@ -0,0 +1,45 @@
+# Test for prologue skipping in minimal symbols with line info when
+# compiled with -O0.  Test workaround gcc bug 49906.
+# Copyright 2011 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+set testfile "prologue"
+set srcfile ${testfile}.c
+set binfile ${objdir}/${subdir}/${testfile}
+
+if  { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable {debug optimize=-O0}] != "" } {
+     untested prologue.exp
+     return -1
+}
+
+gdb_exit
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+gdb_load ${binfile}
+
+if ![runto_main] then {
+   fail "Can't run to main"
+   return 0
+}
+
+set loc_end_of_prologue_workaround [gdb_get_line_number "end of prologue workaround"]
+set loc_end_of_prologue_broken [gdb_get_line_number "end of prologue broken"]
+
+# Test less human-confusing breakpoint location in case weirdness is
+# encountered.
+gdb_test "break f_gccbug49906"\
+	"Breakpoint.*at.*file.*$srcfile,.*line.*$loc_end_of_prologue_workaround\\."\
+	"Workaround gcc bug 49906 worked"
+
Index: gdb/testsuite/gdb.base/prologue.c
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/prologue.c,v
retrieving revision 1.6
diff -u -p -r1.6 prologue.c
--- gdb/testsuite/gdb.base/prologue.c	1 Jan 2011 15:33:42 -0000	1.6
+++ gdb/testsuite/gdb.base/prologue.c	25 Oct 2011 17:29:12 -0000
@@ -16,6 +16,15 @@
    along with this program.  If not, see <http://www.gnu.org/licenses/>.
    */
 
+
+static int i;
+static void
+f_gccbug49906 (void)
+{  for (;;) /* end of prologue workaround  - Weird style intended! */
+     if (i++)
+       break;
+} /* end of prologue broken */
+
 int leaf (void)
 {
   return 1;
@@ -33,6 +42,7 @@ int main(void)
 {
   marker (0);
   marker (0);
+  f_gccbug49906 ();  /* set breakpoint here */
 
   return 0;
 }