2886 Commits

Author SHA1 Message Date
Nicholas Nethercote
26afa60bf2 Update website address.
MERGE TO 2.4 REPOSITORY



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3753
2005-05-17 03:22:38 +00:00
Julian Seward
a548e818d0 A 4999 entry hashtable is too small when there are a huge number (500k
ish) of blocks simultaneously live, and the client is doing intensive
malloc/frees.  Increase the table size 16-fold.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3752
2005-05-17 02:20:35 +00:00
Nicholas Nethercote
937a49ea05 Added module m_options for holding all the command-line option stuff.
Perhaps parts of process_cmd_line_option() should go in here, but I've
not done that for now.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3751
2005-05-16 23:31:24 +00:00
Tom Hughes
81e30f248f Fix test for amd64.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3750
2005-05-16 23:16:25 +00:00
Nicholas Nethercote
09871f18f0 Move VGA_STACK_SIZE_W into m_syscalls, the only place that uses it.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3749
2005-05-16 21:39:23 +00:00
Nicholas Nethercote
df94111968 Removed the core_arch_asm.h files. The OYNK macros now live in tool_asm.h,
at least for the moment.


git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3748
2005-05-16 21:30:24 +00:00
Nicholas Nethercote
9c67331c7d Move ThreadArchState into core.h and avoid duplication.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3747
2005-05-16 21:16:44 +00:00
Nicholas Nethercote
9e20a9de6f Remove VGP_LONGJMP and VGP_SETJMP, since they're the same on all
platforms.


git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3746
2005-05-16 20:50:52 +00:00
Nicholas Nethercote
586e79e4bd core.h: remove some #include lines, move some others to more localised
places.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3745
2005-05-16 20:40:51 +00:00
Julian Seward
dab52189a5 wibble
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3744
2005-05-16 18:30:40 +00:00
Julian Seward
650de479d1 CFI reader: handle DW_CFA_GNU_window_save. Why is this sparc-specific
construct turning up in .so's in x86 SuSE 9.3 ?  Answers on a postcard
please.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3743
2005-05-16 18:25:09 +00:00
Julian Seward
1038a125cb More FC4 jiggery-pokery.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3742
2005-05-16 16:39:57 +00:00
Julian Seward
9ca614ab65 Disable a bunch of suppressions made obsolete by r3740.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3741
2005-05-16 16:39:28 +00:00
Julian Seward
868d021e86 Fix for sys_sysctl wrapper, from Ulrich Drepper.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3740
2005-05-16 16:38:45 +00:00
Nicholas Nethercote
816419fa9a No need to generate valgrind.h from valgrind.h.in. We can just use the
compiler symbols rather than our own symbols for the architectures (we
already were, more or less).  This simplifies the build, hurrah!

I also inverted the sense of the NVALGRIND ifdefs, to make them easier to
read, and fixed up some comments.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3739
2005-05-16 13:31:23 +00:00
Julian Seward
b6de06b6dc CFI reader: partially handle DW_CFA_expression and
DW_CFA_def_cfa_expression in the sense that they are parsed correctly,
but the contained expression is ignored.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3738
2005-05-16 11:48:34 +00:00
Julian Seward
2a2b57fdc5 Fix comment.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3737
2005-05-16 11:47:17 +00:00
Tom Hughes
0cee6603ab Fill in the amd64 case for regaddr_from_tst.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3736
2005-05-16 07:31:57 +00:00
Tom Hughes
c829917c1f Use movq to get the frame pointer on amd64.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3735
2005-05-16 07:31:05 +00:00
Nicholas Nethercote
5f3248aeb3 Complete restructuring of include/ directory -- I forgot about the ARM
stuff.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3734
2005-05-16 03:36:19 +00:00
Nicholas Nethercote
8c63e060a7 Restructured the include/ directory.
- All the subdirectories have gone: arm/, x86/, amd64/, linux/,         
  x86-linux/, amd64-linux/, arm-linux/.

- The following files were moved out of those directories into include/:

    amd64-linux/vki_arch.h            --> vki-amd64-linux.h
    x86-linux/vki_arch.h              --> vki-x86-linux.h
    x86-linux/vki_arch_posixtypes.h   --> vki_posixtypes-x86-linux.h
    linux/vki.h                       --> vki-linux.h
    amd64-linux/vki_arch_posixtypes.h --> vki_posixtypes-amd64-linux.h

- The following very small files were inlined into tool.h using the "#if
  defined(VGP_x86)..." technique:

    x86/tool_arch.h
    arm/tool_arch.h
    amd64/tool_arch.h

  The same technique was used twice to include the appropriate
  vki-$PLATFORM and vki-$OS files into tool.h.  

- The other files in those directories were removed.

- The build is much simpler, since we have 7(!) fewer Makefile.am files.
  Far fewer -I options are needed when compiling, too.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3733
2005-05-16 03:25:12 +00:00
Nicholas Nethercote
1365301e39 Moved the stack-walking macros into m_stacktrace.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3732
2005-05-16 00:16:56 +00:00
Nicholas Nethercote
b1b8637207 Move some decls in vg_dwarf.c closer to the comment explaining them.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3731
2005-05-16 00:04:02 +00:00
Nicholas Nethercote
87429d4894 Moved VGA_(reg_addr_from_tst)() and VGA_R_{STACK,FRAME}_PTR into vg_symtab2.c.
Also made regaddr() and co. 64-bit clean (I hope I did that correctly).





git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3730
2005-05-16 00:01:04 +00:00
Nicholas Nethercote
0fb8967ab3 Move VGA_GET_REAL_{STACK,FRAME}_PTR into vg_symtab2.c.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3729
2005-05-15 23:36:12 +00:00
Nicholas Nethercote
d36924d966 bugfix: don't print bogus "obj:" lines when generating suppressions.
Thanks to Julian for spotting and diagnosing it.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3728
2005-05-15 21:09:40 +00:00
Nicholas Nethercote
e093b9b370 Let's assume the client's address space starts at 0 until we port
to a platform for which that's not true.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3727
2005-05-15 21:03:42 +00:00
Nicholas Nethercote
7a6ee2fc93 Improved structure of LDT-related code:
- one declarations from core.h removed, one moved to within m_syscalls.
- all the x86 LDT stuff made local to m_syscalls.  x86-linux/ldt.c removed
  as a result.  x86/state.c slimmed down, too.  x86/x86_private.h removed
  too.
- all the AMD64 LDT stuff was deleted, since it was all commented out.  It
  can be added back in later in the appropriate places if necessary.
  Thus amd64-linux/ldt.c and amd64/amd64_private.h were removed.
- other minor naming changes

I hope I didn't break AMD64 compilation.


git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3726
2005-05-15 20:52:04 +00:00
Nicholas Nethercote
c14a5bb6a6 Cover the unknown-platform case.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3725
2005-05-15 18:51:47 +00:00
Nicholas Nethercote
6daae32217 Remove dead constant.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3724
2005-05-15 17:38:56 +00:00
Nicholas Nethercote
92c61a5fd1 Don't mix backticks and apostrophes when quoting words -- eg. use 'foo'
rather than `foo', as www.cl.cam.ac.uk/~mgk25/ucs/quotes.html explains
we should (in more detail than you'd imagine was possible).  I did this
both in output messages and in some comments, for consistency.


git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3723
2005-05-15 17:28:26 +00:00
Julian Seward
c17fb2042b glibc-2.3.5 fix (FC4)
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3722
2005-05-15 17:18:40 +00:00
Julian Seward
4ee2e07775 CFI reader: handle DW_CFA_def_cfa_offset_sf -- FC4 needs it
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3721
2005-05-15 17:14:28 +00:00
Nicholas Nethercote
c91f5ac048 Cleaned up the logging command-line option variables:
- renamed VG_(logging_to_filedes) as VG_(logging_to_socket), which is
  clearer
- no longer exporting clo_log_to, which avoids the confusion about whether
  it or VG_(logging_to_socket) actually controls where output goes
- couple of other minor cleanups


git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3720
2005-05-15 16:22:58 +00:00
Nicholas Nethercote
09cb7d57d1 Add note to FAQ about unloaded shared objects and leak errors.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3719
2005-05-15 14:49:24 +00:00
Nicholas Nethercote
3040604b6d Remove VGA_WORD_SIZE; we already have VG_WORDSIZE.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3718
2005-05-15 03:56:26 +00:00
Nicholas Nethercote
944f1aa854 fixup headers a bit
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3717
2005-05-15 03:52:40 +00:00
Nicholas Nethercote
44f350c366 Move the symbol table reading closer to the debug info reading.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3716
2005-05-15 03:38:53 +00:00
Nicholas Nethercote
4570a48d51 Be more concise.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3715
2005-05-15 03:30:56 +00:00
Nicholas Nethercote
0667c3947d Remove call to VGA_(cleanup_thread)() from VG_(exit_thread)(), since
mostly_clear_thread_record() calls it anyway.


git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3714
2005-05-14 23:28:42 +00:00
Nicholas Nethercote
48d4c8b590 Remove dead function.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3713
2005-05-14 23:10:37 +00:00
Nicholas Nethercote
dee67271fb Whoops, a little premature in removing core_asm.h -- still need it
because it #includes tool_asm.h.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3712
2005-05-14 21:51:49 +00:00
Nicholas Nethercote
8e0e03d480 Created pub_core_transtab_asm.h, killing core_asm.h in the process.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3711
2005-05-14 21:44:20 +00:00
Nicholas Nethercote
357303aebf Modularised vg_hashtable.c as m_hashtable.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3710
2005-05-14 21:28:43 +00:00
Nicholas Nethercote
0367213e2d Renamed the field Segment.symtab to Segment.seginfo to avoid confusion
with SegInfo.symtab.  

Also renamed VG_(symtab_{inc,dec}ref)() as VG_(seginfo_{inc,dec}ref)()
for the same reason.

Also renamed various SegInfo variables from "seg" to "si" to avoid
confusion with the many Segment variables called "seg".



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3709
2005-05-14 18:42:26 +00:00
Nicholas Nethercote
a9294cab3f VG_(resolve_redir)() can now be made private to vg_redir.c.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3708
2005-05-14 17:18:12 +00:00
Nicholas Nethercote
aa4560de8a Move resolve_redir_allsegs() to vg_redir.c, the only file that uses it,
and make it private.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3707
2005-05-14 17:11:06 +00:00
Nicholas Nethercote
7d65a9b8b7 Remove a short sequence of code that was:
(a) overly complex (the parens could be added by the sprintf);
(b) buggy (buf is used uninitialised if VG_(get_fnname) fails);
(c) redundant (the VG_(core_panic_at) prints the stack trace anyway).


git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3706
2005-05-14 17:00:25 +00:00
Julian Seward
98aa9b7354 Add test cases for MOVQ on amd64.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3705
2005-05-14 11:18:31 +00:00
Julian Seward
0704bc9fa9 No wonder I couldn't figure out exactly how movq was supposed to
behave ..  no test cases!  Now fixed.



git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3704
2005-05-14 02:08:48 +00:00