[Valgrind-developers] valgrind: r9480 - branches/DARWIN/coregrind/m_debuginfo

[<sv...@va...>]

Author: sewardj
Date: 2009-03-21 17:14:02 +0000 (Sat, 21 Mar 2009)
New Revision: 9480

Log:
A one-character change that (almost) makes 3 more regtests pass: when
looking for the debug_ranges section, use the correct section name, so
the section is actually found.  This stops ML_(new_dwarf3_reader) from
complaining about a nonsense debug_ranges section.  AFAICT from a
brief investigation, after this change, types and location information
for stack-allocated variables is read correctly.  Information on
global variables still appears to be entirely bogus; if I had to guess
this would be because the di->data_avma/svma/bias etc fields have not
been set correctly (probably they have not been set at all).

memcheck/tests/varinfo{2,5,6} would pass after this change, except for
the unrelated issue that the malloc/freed block counts are too high by
7.  Perhaps Darwin's libc doing some extra mallocs we don't know
about?



Modified:
   branches/DARWIN/coregrind/m_debuginfo/readmacho.c


Modified: branches/DARWIN/coregrind/m_debuginfo/readmacho.c
===================================================================
--- branches/DARWIN/coregrind/m_debuginfo/readmacho.c	2009-03-21 16:55:41 UTC (rev 9479)
+++ branches/DARWIN/coregrind/m_debuginfo/readmacho.c	2009-03-21 17:14:02 UTC (rev 9480)
@@ -642,7 +642,7 @@
                       "__DWARF", "__debug_str", &debug_str_sz);
       debug_ranges_img = 
           getsectdata(dw_oimage, dw_n_oimage, 
-                      "__DWARF", "__debug_aranges", &debug_ranges_sz);
+                      "__DWARF", "__debug_ranges", &debug_ranges_sz);
       debug_loc_img = 
           getsectdata(dw_oimage, dw_n_oimage, 
                       "__DWARF", "__debug_loc", &debug_loc_sz);