This is the mail archive of the
elfutils-devel@sourceware.org
mailing list for the elfutils project.
FYI: Make libdw dwarf_child () work for debug_type DIEs
- From: Mark Wielaard <mjw at redhat dot com>
- To: elfutils-devel at lists dot fedorahosted dot org
- Date: Tue, 26 Apr 2011 17:24:36 +0200
- Subject: FYI: Make libdw dwarf_child () work for debug_type DIEs
Hi,
There was a small bug in dwarf_child () that made it fail
with .debug_type Dwarf_Dies. It contained a sanity check to see whether
we reached the end of the section early. A Dwarf_CU can come from either
the debug_info unit or a debug_type unit section. So use the correct one
for sanity checking whether we have a truncated section. Fixed as
follows:
2011-04-26 Mark Wielaard <mjw@redhat.com>
* dwarf_child (dwarf_child): Sanity check end of section against
cu_data () of die->cu.
Cheers,
Mark
commit f4c39c7ccba57bcc71ab92760f4194b3e4667bb7
Author: Mark Wielaard <mjw@redhat.com>
Date: Tue Apr 26 17:15:28 2011 +0200
libdw: Sanity check end of section against cu_data () of die->cu.
A Dwarf_CU can come from either the debug_info compile unit or a
debug_type type unit section. Use the correct one for sanity checking
whether we have a truncated section.
diff --git a/libdw/ChangeLog b/libdw/ChangeLog
index 0b17461..8febd42 100644
--- a/libdw/ChangeLog
+++ b/libdw/ChangeLog
@@ -1,3 +1,8 @@
+2011-04-26 Mark Wielaard <mjw@redhat.com>
+
+ * dwarf_child (dwarf_child): Sanity check end of section against
+ cu_data () of die->cu.
+
2011-03-22 Mark Wielaard <mjw@redhat.com>
* dwarf.h: Add DW_TAG_GNU_call_site,
diff --git a/libdw/dwarf_child.c b/libdw/dwarf_child.c
index 1ec3704..0c2df00 100644
--- a/libdw/dwarf_child.c
+++ b/libdw/dwarf_child.c
@@ -177,13 +177,13 @@ dwarf_child (die, result)
/* RESULT can be the same as DIE. So preserve what we need. */
struct Dwarf_CU *cu = die->cu;
+ Elf_Data *cu_sec = cu_data (cu);
/* It's kosher (just suboptimal) to have a null entry first thing (7.5.3).
So if this starts with ULEB128 of 0 (even with silly encoding of 0),
it is a kosher null entry and we do not really have any children. */
const unsigned char *code = addr;
- const unsigned char *endp = (cu->dbg->sectiondata[IDX_debug_info]->d_buf
- + cu->dbg->sectiondata[IDX_debug_info]->d_size);
+ const unsigned char *endp = (cu_sec->d_buf + cu_sec->d_size);
while (1)
{
if (unlikely (code >= endp)) /* Truncated section. */