This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[patch] Regression: Re: FYI: remove free_symtab


Hi Tom,

heavy crash regression on Fedora 15, in fact on any system using .gdb_index.
One of such testfiles is for example gdb.base/cursal.exp.

3405069491ed7249b28958123166f160d7fe7fcb is the first bad commit
commit 3405069491ed7249b28958123166f160d7fe7fcb
Author: Tom Tromey <tromey@redhat.com>
Date:   Mon Apr 4 15:19:54 2011 +0000

        * symtab.h (free_symtab): Remove.
        (forget_cached_source_info_for_objfile): Declare.
        * symmisc.c (free_symtab): Remove.
        * source.c (forget_cached_source_info_for_objfile): New function.
        (forget_cached_source_info): Use it.
        * objfiles.c (free_objfile): Simplify check before calling
        clear_current_source_symtab_and_line.  Call
        forget_cached_source_info_for_objfile.

:040000 040000 be020dc3c59fd93e1e2dd2372227733c585a9e83 2ab4f20b502413ce56eafe87828ae10f7f7f1ee8 M      gdb
bisect run success

Program received signal SIGSEGV, Segmentation fault.
0x0000000000488b0e in objfile_data (objfile=0x2128400, data=0x1e25a20) at objfiles.c:1480
1480      return objfile->data[data->index];
(gdb) p *objfile
$1 = {next = 0x0, name = 0x20fdf60 "/usr/lib/debug/lib64/ld-2.13.90.so.debug",

objfile->data is already NULL.

#0  in objfile_data (objfile=0x2128400, data=0x1e25a20) at objfiles.c:1480
#1  in dw2_setup (objfile=0x2128400) at dwarf2read.c:2139
#2  in dw2_forget_cached_source_info (objfile=0x2128400) at dwarf2read.c:2308
#3  in forget_cached_source_info_for_objfile (objfile=0x2128400) at source.c:360
#4  in free_objfile (objfile=0x2128400) at objfiles.c:640
#5  in objfile_purge_solibs () at objfiles.c:988
#6  in no_shared_libraries (ignored=0x0, from_tty=0) at solib.c:1308
#7  in target_pre_inferior (from_tty=0) at target.c:2390
#8  in run_command_1 (args=0x0, from_tty=0, tbreak_at_main=0) at infcmd.c:506
#9  in run_command (args=0x0, from_tty=0) at infcmd.c:610

I do not see any deep design in the freeing order so I have just reshuffled it
so that it started working again.

No regressions on {x86_64,x86_64-m32,i686}-fedora{13,14,15,-rawhide}-linux-gnu
together with the debugformat (as has been posted along) fix against yesterday.

I will check it in today.


Thanks,
Jan


gdb/
2011-04-05  Jan Kratochvil  <jan.kratochvil@redhat.com>

	Fix crash regression on systems featuring .gdb_index.
	* objfiles.c (free_objfile): Move the
	forget_cached_source_info_for_objfile call earlier.  Comment it.
	Extend the comment for objfile_free_data.

--- a/gdb/objfiles.c
+++ b/gdb/objfiles.c
@@ -583,6 +583,10 @@ free_objfile (struct objfile *objfile)
      lists.  */
   preserve_values (objfile);
 
+  /* It still may reference data modules have associated with the objfile and
+     the symbol file data.  */
+  forget_cached_source_info_for_objfile (objfile);
+
   /* First do any symbol file specific actions required when we are
      finished with a particular symbol file.  Note that if the objfile
      is using reusable symbol information (via mmalloc) then each of
@@ -595,7 +599,8 @@ free_objfile (struct objfile *objfile)
       (*objfile->sf->sym_finish) (objfile);
     }
 
-  /* Discard any data modules have associated with the objfile.  */
+  /* Discard any data modules have associated with the objfile.  The function
+     still may reference objfile->obfd.  */
   objfile_free_data (objfile);
 
   gdb_bfd_unref (objfile->obfd);
@@ -637,8 +642,6 @@ free_objfile (struct objfile *objfile)
       clear_current_source_symtab_and_line ();
   }
 
-  forget_cached_source_info_for_objfile (objfile);
-
   /* The last thing we do is free the objfile struct itself.  */
 
   xfree (objfile->name);


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]