[Bug symtab/26092] Complain about contradictory DW_LNE_end_sequence marker

["vries at gcc dot gnu.org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=26092

--- Comment #1 from Tom de Vries <vries at gcc dot gnu.org> ---
I wrote a tentative patch detecting this problem using a gdb_assert:
...
diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
index 477c382b81..1219cc58be 100644
--- a/gdb/dwarf2/read.c
+++ b/gdb/dwarf2/read.c
@@ -20045,6 +20045,7 @@ class lnp_state_machine
   /* These are initialized in the constructor.  */

   CORE_ADDR m_address;
+  CORE_ADDR m_prev_line_address = 0;
   bool m_is_stmt;
   unsigned int m_discriminator;

@@ -20235,6 +20236,19 @@ dwarf_finish_line (struct gdbarch *gdbarch, struct
subfile *subfile,
 void
 lnp_state_machine::record_line (bool end_sequence)
 {
+  if (end_sequence)
+    {
+      if (m_prev_line_address != 0)
+       {
+         gdb_assert (m_prev_line_address != m_address);
+         m_prev_line_address = 0;
+       }
+    }
+  else
+    {
+      m_prev_line_address = m_address;
+    }
+
   if (dwarf_line_debug)
     {
       fprintf_unfiltered (gdb_stdlog,
...

I did a gdb.dwarf2 testsuite run, hoping to find that the problem is only
detected in the dwarf assembly test-cases.

However, it also triggered on this bit in libgcc:
...
  [0x0000064a]  Special opcode 75: advance Address by 5 to 0x300a and Line by 0
to 246
  [0x0000064b]  Advance PC by 0 to 0x300a
  [0x0000064d]  Extended opcode 1: End of Sequence
...
So it looks like we also got compilers/assemblers generating this.

-- 
You are receiving this mail because:
You are on the CC list for the bug.