NEWS

Release 3.3.0 (XX XXX 2007)
~~~~~~~~~~~~~~~~~~~~~~~~~~~
3.3.0 is a feature release with ...

XXX: AIX support (ppc32 and ppc64?)

Other user-visible changes:

- A new suppression kind has been introduced: "Jump".  This is for
  suppressing jump-to-invalid-address errors.  Previously you had to use an
  "Addr1" suppression, which didn't make much sense.

- The behaviour of Memcheck's client requests VALGRIND_GET_VBITS and
  VALGRIND_SET_VBITS have changed slightly.  They no longer issue
  addressability errors -- if either array is partially unaddressable,
  they just return 3 (as before).  Also, SET_VBITS doesn't report
  definedness errors if any of the V bits are undefined.

- The behaviour of the tool Lackey has changed slightly.  First, the output
  from --trace-mem has been made more compact, to reduce the size of the
  traces.  Second, a new option --trace-superblocks has been added, which
  shows the addresses of superblocks (code blocks) as they are executed.

Developer-visible changes:

- The names of some functions and types within the Vex IR have changed.
  So any existing tools will have to be updated to reflect these
  changes.  The new names should be clearer.  The file
  VEX/pub/libvex_ir.h is also much better commented, and serves as the
  best documentation about Vex.


Release 3.2.3 (29 Jan 2007)
~~~~~~~~~~~~~~~~~~~~~~~~~~~
Unfortunately 3.2.2 introduced a regression which can cause an
assertion failure ("vex: the `impossible' happened: eqIRConst") when
running obscure pieces of SSE code.  3.2.3 fixes this and adds one
more glibc-2.5 intercept.  In all other respects it is identical to
3.2.2.  Please do not use (or package) 3.2.2; instead use 3.2.3.

n-i-bz   vex: the `impossible' happened: eqIRConst
n-i-bz   Add an intercept for glibc-2.5 __stpcpy_chk

(3.2.3: 29 Jan 2007, vex r1732, valgrind r6560).


Release 3.2.2 (22 Jan 2007)
~~~~~~~~~~~~~~~~~~~~~~~~~~~
3.2.2 fixes a bunch of bugs in 3.2.1, adds support for glibc-2.5 based
systems (openSUSE 10.2, Fedora Core 6), improves support for icc-9.X
compiled code, and brings modest performance improvements in some
areas, including amd64 floating point, powerpc support, and startup
responsiveness on all targets.

The fixed bugs are as follows.  Note that "n-i-bz" stands for "not in
bugzilla" -- that is, a bug that was reported to us but never got a
bugzilla entry.  We encourage you to file bugs in bugzilla
(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
developers (or mailing lists) directly.

129390   ppc?->IR: some kind of VMX prefetch (dstt)
129968   amd64->IR: 0xF 0xAE 0x0 (fxsave)
134319   ==129968
133054   'make install' fails with syntax errors
118903   ==133054
132998   startup fails in when running on UML
134207   pkg-config output contains @VG_PLATFORM@
134727   valgrind exits with "Value too large for defined data type"
n-i-bz   ppc32/64: support mcrfs
n-i-bz   Cachegrind/Callgrind: Update cache parameter detection
135012   x86->IR: 0xD7 0x8A 0xE0 0xD0 (xlat)
125959   ==135012
126147   x86->IR: 0xF2 0xA5 0xF 0x77 (repne movsw)
136650   amd64->IR: 0xC2 0x8 0x0
135421   x86->IR: unhandled Grp5(R) case 6
n-i-bz   Improved documentation of the IR intermediate representation
n-i-bz   jcxz (x86) (users list, 8 Nov)
n-i-bz   ExeContext hashing fix
n-i-bz   fix CFI reading failures ("Dwarf CFI 0:24 0:32 0:48 0:7")
n-i-bz   fix Cachegrind/Callgrind simulation bug
n-i-bz   libmpiwrap.c: fix handling of MPI_LONG_DOUBLE
n-i-bz   make User errors suppressible
136844   corrupted malloc line when using --gen-suppressions=yes
138507   ==136844
n-i-bz   Speed up the JIT's register allocator
n-i-bz   Fix confusing leak-checker flag hints
n-i-bz   Support recent autoswamp versions
n-i-bz   ppc32/64 dispatcher speedups
n-i-bz   ppc64 front end rld/rlw improvements
n-i-bz   ppc64 back end imm64 improvements
136300   support 64K pages on ppc64-linux
139124   == 136300
n-i-bz   fix ppc insn set tests for gcc >= 4.1
137493   x86->IR: recent binutils no-ops
137714   x86->IR: 0x66 0xF 0xF7 0xC6 (maskmovdqu)
138424   "failed in UME with error 22" (produce a better error msg)
138856   ==138424
138627   Enhancement support for prctl ioctls
138896   Add support for usb ioctls
136059   ==138896
139050   ppc32->IR: mfspr 268/269 instructions not handled
n-i-bz   ppc32->IR: lvxl/stvxl
n-i-bz   glibc-2.5 support
n-i-bz   memcheck: provide replacement for mempcpy
n-i-bz   memcheck: replace bcmp in ld.so
n-i-bz   Use 'ifndef' in VEX's Makefile correctly
n-i-bz   Suppressions for MVL 4.0.1 on ppc32-linux
n-i-bz   libmpiwrap.c: Fixes for MPICH
n-i-bz   More robust handling of hinted client mmaps
139776   Invalid read in unaligned memcpy with Intel compiler v9
n-i-bz   Generate valid XML even for very long fn names
n-i-bz   Don't prompt about suppressions for unshown reachable leaks
139910   amd64 rcl is not supported
n-i-bz   DWARF CFI reader: handle DW_CFA_undefined
n-i-bz   DWARF CFI reader: handle icc9 generated CFI info better
n-i-bz   fix false uninit-value errs in icc9 generated FP code
n-i-bz   reduce extraneous frames in libmpiwrap.c
n-i-bz   support pselect6 on amd64-linux

(3.2.2: 22 Jan 2007, vex r1729, valgrind r6545).


Release 3.2.1 (16 Sept 2006)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3.2.1 adds x86/amd64 support for all SSE3 instructions except monitor
and mwait, further reduces memcheck's false error rate on all
platforms, adds support for recent binutils (in OpenSUSE 10.2 and
Fedora Rawhide) and fixes a bunch of bugs in 3.2.0.  Some of the fixed
bugs were causing large programs to segfault with --tool=callgrind and
--tool=cachegrind, so an upgrade is recommended.

In view of the fact that any 3.3.0 release is unlikely to happen until
well into 1Q07, we intend to keep the 3.2.X line alive for a while
yet, and so we tentatively plan a 3.2.2 release sometime in December
06.

The fixed bugs are as follows.  Note that "n-i-bz" stands for "not in
bugzilla" -- that is, a bug that was reported to us but never got a
bugzilla entry.

n-i-bz   Expanding brk() into last available page asserts
n-i-bz   ppc64-linux stack RZ fast-case snafu
n-i-bz   'c' in --gen-supps=yes doesn't work
n-i-bz   VG_N_SEGMENTS too low (users, 28 June)
n-i-bz   VG_N_SEGNAMES too low (Stu Robinson)
106852   x86->IR: fisttp (SSE3)
117172   FUTEX_WAKE does not use uaddr2
124039   Lacks support for VKI_[GP]IO_UNIMAP*
127521   amd64->IR: 0xF0 0x48 0xF 0xC7 (cmpxchg8b)
128917   amd64->IR: 0x66 0xF 0xF6 0xC4 (psadbw,SSE2)
129246   JJ: ppc32/ppc64 syscalls, w/ patch
129358   x86->IR: fisttpl (SSE3)
129866   cachegrind/callgrind causes executable to die
130020   Can't stat .so/.exe error while reading symbols
130388   Valgrind aborts when process calls malloc_trim()
130638   PATCH: ppc32 missing system calls
130785   amd64->IR: unhandled instruction "pushfq"
131481:  (HINT_NOP) vex x86->IR: 0xF 0x1F 0x0 0xF
131298   ==131481
132146   Programs with long sequences of bswap[l,q]s
132918   vex amd64->IR: 0xD9 0xF8 (fprem)
132813   Assertion at priv/guest-x86/toIR.c:652 fails
133051   'cfsi->len > 0 && cfsi->len < 2000000' failed
132722   valgrind header files are not standard C
n-i-bz   Livelocks entire machine (users list, Timothy Terriberry)
n-i-bz   Alex Bennee mmap problem (9 Aug)
n-i-bz   BartV: Don't print more lines of a stack-trace than were obtained.
n-i-bz   ppc32 SuSE 10.1 redir
n-i-bz   amd64 padding suppressions
n-i-bz   amd64 insn printing fix.
n-i-bz   ppc cmp reg,reg fix
n-i-bz   x86/amd64 iropt e/rflag reduction rules
n-i-bz   SuSE 10.1 (ppc32) minor fixes
133678   amd64->IR: 0x48 0xF 0xC5 0xC0 (pextrw?)
133694   aspacem assertion: aspacem_minAddr <= holeStart
n-i-bz   callgrind: fix warning about malformed creator line 
n-i-bz   callgrind: fix annotate script for data produced with 
         --dump-instr=yes
n-i-bz   callgrind: fix failed assertion when toggling 
         instrumentation mode
n-i-bz   callgrind: fix annotate script fix warnings with
         --collect-jumps=yes
n-i-bz   docs path hardwired (Dennis Lubert)

The following bugs were not fixed, due primarily to lack of developer
time, and also because bug reporters did not answer requests for
feedback in time for the release:

129390   ppc?->IR: some kind of VMX prefetch (dstt)
129968   amd64->IR: 0xF 0xAE 0x0 (fxsave)
133054   'make install' fails with syntax errors
n-i-bz   Signal race condition (users list, 13 June, Johannes Berg)
n-i-bz   Unrecognised instruction at address 0x70198EC2 (users list,
         19 July, Bennee)
132998   startup fails in when running on UML

The following bug was tentatively fixed on the mainline but the fix
was considered too risky to push into 3.2.X:

133154   crash when using client requests to register/deregister stack

(3.2.1: 16 Sept 2006, vex r1658, valgrind r6070).


Release 3.2.0 (7 June 2006)
~~~~~~~~~~~~~~~~~~~~~~~~~~~
3.2.0 is a feature release with many significant improvements and the
usual collection of bug fixes.  This release supports X86/Linux,
AMD64/Linux, PPC32/Linux and PPC64/Linux.

Performance, especially of Memcheck, is improved, Addrcheck has been
removed, Callgrind has been added, PPC64/Linux support has been added,
Lackey has been improved, and MPI support has been added.  In detail:

- Memcheck has improved speed and reduced memory use.  Run times are
  typically reduced by 15-30%, averaging about 24% for SPEC CPU2000.
  The other tools have smaller but noticeable speed improvments.  We
  are interested to hear what improvements users get.

  Memcheck uses less memory due to the introduction of a compressed
  representation for shadow memory.  The space overhead has been
  reduced by a factor of up to four, depending on program behaviour.
  This means you should be able to run programs that use more memory
  than before without hitting problems.

- Addrcheck has been removed.  It has not worked since version 2.4.0,
  and the speed and memory improvements to Memcheck make it redundant.
  If you liked using Addrcheck because it didn't give undefined value
  errors, you can use the new Memcheck option --undef-value-errors=no
  to get the same behaviour.

- The number of undefined-value errors incorrectly reported by
  Memcheck has been reduced (such false reports were already very
  rare).  In particular, efforts have been made to ensure Memcheck
  works really well with gcc 4.0/4.1-generated code on X86/Linux and
  AMD64/Linux.

- Josef Weidendorfer's popular Callgrind tool has been added.  Folding
  it in was a logical step given its popularity and usefulness, and
  makes it easier for us to ensure it works "out of the box" on all
  supported targets.  The associated KDE KCachegrind GUI remains a
  separate project.

- A new release of the Valkyrie GUI for Memcheck, version 1.2.0,
  accompanies this release.  Improvements over previous releases
  include improved robustness, many refinements to the user interface,
  and use of a standard autoconf/automake build system.  You can get
  it from http://www.valgrind.org/downloads/guis.html.

- Valgrind now works on PPC64/Linux.  As with the AMD64/Linux port,
  this supports programs using to 32G of address space.  On 64-bit
  capable PPC64/Linux setups, you get a dual architecture build so
  that both 32-bit and 64-bit executables can be run.  Linux on POWER5
  is supported, and POWER4 is also believed to work.  Both 32-bit and
  64-bit DWARF2 is supported.  This port is known to work well with
  both gcc-compiled and xlc/xlf-compiled code.

- Floating point accuracy has been improved for PPC32/Linux.
  Specifically, the floating point rounding mode is observed on all FP
  arithmetic operations, and multiply-accumulate instructions are
  preserved by the compilation pipeline.  This means you should get FP
  results which are bit-for-bit identical to a native run.  These
  improvements are also present in the PPC64/Linux port.

- Lackey, the example tool, has been improved:

  * It has a new option --detailed-counts (off by default) which
    causes it to print out a count of loads, stores and ALU operations
    done, and their sizes.

  * It has a new option --trace-mem (off by default) which causes it
    to print out a trace of all memory accesses performed by a
    program.  It's a good starting point for building Valgrind tools
    that need to track memory accesses.  Read the comments at the top
    of the file lackey/lk_main.c for details.

  * The original instrumentation (counting numbers of instructions,
    jumps, etc) is now controlled by a new option --basic-counts.  It
    is on by default.

- MPI support: partial support for debugging distributed applications
  using the MPI library specification has been added.  Valgrind is
  aware of the memory state changes caused by a subset of the MPI
  functions, and will carefully check data passed to the (P)MPI_
  interface.

- A new flag, --error-exitcode=, has been added.  This allows changing
  the exit code in runs where Valgrind reported errors, which is
  useful when using Valgrind as part of an automated test suite.

- Various segfaults when reading old-style "stabs" debug information
  have been fixed.

- A simple performance evaluation suite has been added.  See
  perf/README and README_DEVELOPERS for details.  There are
  various bells and whistles.

- New configuration flags:
    --enable-only32bit
    --enable-only64bit
  By default, on 64 bit platforms (ppc64-linux, amd64-linux) the build
  system will attempt to build a Valgrind which supports both 32-bit
  and 64-bit executables.  This may not be what you want, and you can
  override the default behaviour using these flags.

Please note that Helgrind is still not working.  We have made an
important step towards making it work again, however, with the
addition of function wrapping (see below).

Other user-visible changes:

- Valgrind now has the ability to intercept and wrap arbitrary
  functions.  This is a preliminary step towards making Helgrind work
  again, and was required for MPI support.

- There are some changes to Memcheck's client requests.  Some of them
  have changed names:

    MAKE_NOACCESS  --> MAKE_MEM_NOACCESS
    MAKE_WRITABLE  --> MAKE_MEM_UNDEFINED
    MAKE_READABLE  --> MAKE_MEM_DEFINED

    CHECK_WRITABLE --> CHECK_MEM_IS_ADDRESSABLE
    CHECK_READABLE --> CHECK_MEM_IS_DEFINED
    CHECK_DEFINED  --> CHECK_VALUE_IS_DEFINED

  The reason for the change is that the old names are subtly
  misleading.  The old names will still work, but they are deprecated
  and may be removed in a future release.

  We also added a new client request:
  
    MAKE_MEM_DEFINED_IF_ADDRESSABLE(a, len)
    
  which is like MAKE_MEM_DEFINED but only affects a byte if the byte is
  already addressable.

- The way client requests are encoded in the instruction stream has
  changed.  Unfortunately, this means 3.2.0 will not honour client
  requests compiled into binaries using headers from earlier versions
  of Valgrind.  We will try to keep the client request encodings more 
  stable in future.

BUGS FIXED:

108258   NPTL pthread cleanup handlers not called 
117290   valgrind is sigKILL'd on startup
117295   == 117290
118703   m_signals.c:1427 Assertion 'tst->status == VgTs_WaitSys'
118466   add %reg, %reg generates incorrect validity for bit 0
123210   New: strlen from ld-linux on amd64
123244   DWARF2 CFI reader: unhandled CFI instruction 0:18
123248   syscalls in glibc-2.4: openat, fstatat, symlinkat
123258   socketcall.recvmsg(msg.msg_iov[i] points to uninit
123535   mremap(new_addr) requires MREMAP_FIXED in 4th arg
123836   small typo in the doc
124029   ppc compile failed: `vor' gcc 3.3.5
124222   Segfault: @@don't know what type ':' is
124475   ppc32: crash (syscall?) timer_settime()
124499   amd64->IR: 0xF 0xE 0x48 0x85 (femms)
124528   FATAL: aspacem assertion failed: segment_is_sane
124697   vex x86->IR: 0xF 0x70 0xC9 0x0 (pshufw)
124892   vex x86->IR: 0xF3 0xAE (REPx SCASB)
126216   == 124892
124808   ppc32: sys_sched_getaffinity() not handled
n-i-bz   Very long stabs strings crash m_debuginfo
n-i-bz   amd64->IR: 0x66 0xF 0xF5 (pmaddwd)
125492   ppc32: support a bunch more syscalls
121617   ppc32/64: coredumping gives assertion failure
121814   Coregrind return error as exitcode patch
126517   == 121814
125607   amd64->IR: 0x66 0xF 0xA3 0x2 (btw etc)
125651   amd64->IR: 0xF8 0x49 0xFF 0xE3 (clc?)
126253   x86 movx is wrong
126451   3.2 SVN doesn't work on ppc32 CPU's without FPU
126217   increase # threads
126243   vex x86->IR: popw mem
126583   amd64->IR: 0x48 0xF 0xA4 0xC2 (shld $1,%rax,%rdx)
126668   amd64->IR: 0x1C 0xFF (sbb $0xff,%al)
126696   support for CDROMREADRAW ioctl and CDROMREADTOCENTRY fix
126722   assertion: segment_is_sane at m_aspacemgr/aspacemgr.c:1624
126938   bad checking for syscalls linkat, renameat, symlinkat

(3.2.0RC1: 27 May  2006, vex r1626, valgrind r5947).
(3.2.0:     7 June 2006, vex r1628, valgrind r5957).


Release 3.1.1 (15 March 2006)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3.1.1 fixes a bunch of bugs reported in 3.1.0.  There is no new
functionality.  The fixed bugs are:

(note: "n-i-bz" means "not in bugzilla" -- this bug does not have
 a bugzilla entry).

n-i-bz   ppc32: fsub 3,3,3 in dispatcher doesn't clear NaNs
n-i-bz   ppc32: __NR_{set,get}priority
117332   x86: missing line info with icc 8.1
117366   amd64: 0xDD 0x7C fnstsw
118274   == 117366
117367   amd64: 0xD9 0xF4 fxtract
117369   amd64: __NR_getpriority (140)
117419   ppc32: lfsu f5, -4(r11)
117419   ppc32: fsqrt
117936   more stabs problems (segfaults while reading debug info)
119914   == 117936
120345   == 117936
118239   amd64: 0xF 0xAE 0x3F (clflush)
118939   vm86old system call
n-i-bz   memcheck/tests/mempool reads freed memory
n-i-bz   AshleyP's custom-allocator assertion
n-i-bz   Dirk strict-aliasing stuff
n-i-bz   More space for debugger cmd line (Dan Thaler)
n-i-bz   Clarified leak checker output message
n-i-bz   AshleyP's --gen-suppressions output fix
n-i-bz   cg_annotate's --sort option broken
n-i-bz   OSet 64-bit fastcmp bug
n-i-bz   VG_(getgroups) fix (Shinichi Noda)
n-i-bz   ppc32: allocate from callee-saved FP/VMX regs
n-i-bz   misaligned path word-size bug in mc_main.c
119297   Incorrect error message for sse code
120410   x86: prefetchw (0xF 0xD 0x48 0x4)
120728   TIOCSERGETLSR, TIOCGICOUNT, HDIO_GET_DMA ioctls
120658   Build fixes for gcc 2.96
120734   x86: Support for changing EIP in signal handler
n-i-bz   memcheck/tests/zeropage de-looping fix
n-i-bz   x86: fxtract doesn't work reliably
121662   x86: lock xadd (0xF0 0xF 0xC0 0x2)
121893   calloc does not always return zeroed memory
121901   no support for syscall tkill
n-i-bz   Suppression update for Debian unstable
122067   amd64: fcmovnu (0xDB 0xD9)
n-i-bz   ppc32: broken signal handling in cpu feature detection
n-i-bz   ppc32: rounding mode problems (improved, partial fix only)
119482   ppc32: mtfsb1
n-i-bz   ppc32: mtocrf/mfocrf

(3.1.1:  15 March 2006, vex r1597, valgrind r5771).


Release 3.1.0 (25 November 2005)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3.1.0 is a feature release with a number of significant improvements:
AMD64 support is much improved, PPC32 support is good enough to be
usable, and the handling of memory management and address space is
much more robust.  In detail:

- AMD64 support is much improved.  The 64-bit vs. 32-bit issues in
  3.0.X have been resolved, and it should "just work" now in all
  cases.  On AMD64 machines both 64-bit and 32-bit versions of
  Valgrind are built.  The right version will be invoked
  automatically, even when using --trace-children and mixing execution
  between 64-bit and 32-bit executables.  Also, many more instructions
  are supported.

- PPC32 support is now good enough to be usable.  It should work with
  all tools, but please let us know if you have problems.  Three
  classes of CPUs are supported: integer only (no FP, no Altivec),
  which covers embedded PPC uses, integer and FP but no Altivec
  (G3-ish), and CPUs capable of Altivec too (G4, G5).

- Valgrind's address space management has been overhauled.  As a
  result, Valgrind should be much more robust with programs that use
  large amounts of memory.  There should be many fewer "memory
  exhausted" messages, and debug symbols should be read correctly on
  large (eg. 300MB+) executables.  On 32-bit machines the full address
  space available to user programs (usually 3GB or 4GB) can be fully
  utilised.  On 64-bit machines up to 32GB of space is usable; when
  using Memcheck that means your program can use up to about 14GB.

  A side effect of this change is that Valgrind is no longer protected
  against wild writes by the client.  This feature was nice but relied
  on the x86 segment registers and so wasn't portable.

- Most users should not notice, but as part of the address space
  manager change, the way Valgrind is built has been changed.  Each
  tool is now built as a statically linked stand-alone executable,
  rather than as a shared object that is dynamically linked with the
  core.  The "valgrind" program invokes the appropriate tool depending
  on the --tool option.  This slightly increases the amount of disk
  space used by Valgrind, but it greatly simplified many things and
  removed Valgrind's dependence on glibc.

Please note that Addrcheck and Helgrind are still not working.  Work
is underway to reinstate them (or equivalents).  We apologise for the
inconvenience.

Other user-visible changes:

- The --weird-hacks option has been renamed --sim-hints.

- The --time-stamp option no longer gives an absolute date and time.
  It now prints the time elapsed since the program began.

- It should build with gcc-2.96.

- Valgrind can now run itself (see README_DEVELOPERS for how).
  This is not much use to you, but it means the developers can now
  profile Valgrind using Cachegrind.  As a result a couple of
  performance bad cases have been fixed.

- The XML output format has changed slightly.  See
  docs/internals/xml-output.txt.

- Core dumping has been reinstated (it was disabled in 3.0.0 and 3.0.1).
  If your program crashes while running under Valgrind, a core file with
  the name "vgcore.<pid>" will be created (if your settings allow core
  file creation).  Note that the floating point information is not all
  there.  If Valgrind itself crashes, the OS will create a normal core
  file.

The following are some user-visible changes that occurred in earlier
versions that may not have been announced, or were announced but not
widely noticed.  So we're mentioning them now.

- The --tool flag is optional once again;  if you omit it, Memcheck
  is run by default.

- The --num-callers flag now has a default value of 12.  It was
  previously 4.

- The --xml=yes flag causes Valgrind's output to be produced in XML
  format.  This is designed to make it easy for other programs to
  consume Valgrind's output.  The format is described in the file
  docs/internals/xml-format.txt.

- The --gen-suppressions flag supports an "all" value that causes every
  suppression to be printed without asking.

- The --log-file option no longer puts "pid" in the filename, eg. the
  old name "foo.pid12345" is now "foo.12345".

- There are several graphical front-ends for Valgrind, such as Valkyrie,
  Alleyoop and Valgui.  See http://www.valgrind.org/downloads/guis.html
  for a list.

BUGS FIXED:

109861  amd64 hangs at startup
110301  ditto
111554  valgrind crashes with Cannot allocate memory
111809  Memcheck tool doesn't start java
111901  cross-platform run of cachegrind fails on opteron
113468  (vgPlain_mprotect_range): Assertion 'r != -1' failed.
 92071  Reading debugging info uses too much memory
109744  memcheck loses track of mmap from direct ld-linux.so.2
110183  tail of page with _end
 82301  FV memory layout too rigid
 98278  Infinite recursion possible when allocating memory
108994  Valgrind runs out of memory due to 133x overhead
115643  valgrind cannot allocate memory
105974  vg_hashtable.c static hash table
109323  ppc32: dispatch.S uses Altivec insn, which doesn't work on POWER. 
109345  ptrace_setregs not yet implemented for ppc
110831  Would like to be able to run against both 32 and 64 bit 
        binaries on AMD64
110829  == 110831
111781  compile of valgrind-3.0.0 fails on my linux (gcc 2.X prob)
112670  Cachegrind: cg_main.c:486 (handleOneStatement ...
112941  vex x86: 0xD9 0xF4 (fxtract)
110201  == 112941
113015  vex amd64->IR: 0xE3 0x14 0x48 0x83 (jrcxz)
113126  Crash with binaries built with -gstabs+/-ggdb
104065  == 113126
115741  == 113126
113403  Partial SSE3 support on x86
113541  vex: Grp5(x86) (alt encoding inc/dec) case 1
113642  valgrind crashes when trying to read debug information
113810  vex x86->IR: 66 0F F6 (66 + PSADBW == SSE PSADBW)
113796  read() and write() do not work if buffer is in shared memory
113851  vex x86->IR: (pmaddwd): 0x66 0xF 0xF5 0xC7
114366  vex amd64 cannnot handle __asm__( "fninit" )
114412  vex amd64->IR: 0xF 0xAD 0xC2 0xD3 (128-bit shift, shrdq?)
114455  vex amd64->IR: 0xF 0xAC 0xD0 0x1 (also shrdq)
115590: amd64->IR: 0x67 0xE3 0x9 0xEB (address size override)
115953  valgrind svn r5042 does not build with parallel make (-j3)
116057  maximum instruction size - VG_MAX_INSTR_SZB too small?
116483  shmat failes with invalid argument
102202  valgrind crashes when realloc'ing until out of memory
109487  == 102202
110536  == 102202
112687  == 102202
111724  vex amd64->IR: 0x41 0xF 0xAB (more BT{,S,R,C} fun n games)
111748  vex amd64->IR: 0xDD 0xE2 (fucom)
111785  make fails if CC contains spaces
111829  vex x86->IR: sbb AL, Ib
111851  vex x86->IR: 0x9F 0x89 (lahf/sahf)
112031  iopl on AMD64 and README_MISSING_SYSCALL_OR_IOCTL update
112152  code generation for Xin_MFence on x86 with SSE0 subarch
112167  == 112152
112789  == 112152
112199  naked ar tool is used in vex makefile
112501  vex x86->IR: movq (0xF 0x7F 0xC1 0xF) (mmx MOVQ)
113583  == 112501
112538  memalign crash
113190  Broken links in docs/html/
113230  Valgrind sys_pipe on x86-64 wrongly thinks file descriptors
        should be 64bit
113996  vex amd64->IR: fucomp (0xDD 0xE9)
114196  vex x86->IR: out %eax,(%dx) (0xEF 0xC9 0xC3 0x90)
114289  Memcheck fails to intercept malloc when used in an uclibc environment
114756  mbind syscall support
114757  Valgrind dies with assertion: Assertion 'noLargerThan > 0' failed
114563  stack tracking module not informed when valgrind switches threads
114564  clone() and stacks
114565  == 114564
115496  glibc crashes trying to use sysinfo page
116200  enable fsetxattr, fgetxattr, and fremovexattr for amd64

(3.1.0RC1: 20 November 2005, vex r1466, valgrind r5224).
(3.1.0:    26 November 2005, vex r1471, valgrind r5235).


Release 3.0.1 (29 August 2005)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3.0.1 fixes a bunch of bugs reported in 3.0.0.  There is no new
functionality.  Some of the fixed bugs are critical, so if you
use/distribute 3.0.0, an upgrade to 3.0.1 is recommended.  The fixed
bugs are:

(note: "n-i-bz" means "not in bugzilla" -- this bug does not have
 a bugzilla entry).

109313  (== 110505) x86 cmpxchg8b
n-i-bz  x86: track but ignore changes to %eflags.AC (alignment check)
110102  dis_op2_E_G(amd64)
110202  x86 sys_waitpid(#286)
110203  clock_getres(,0)
110208  execve fail wrong retval
110274  SSE1 now mandatory for x86
110388  amd64 0xDD 0xD1
110464  amd64 0xDC 0x1D FCOMP
110478  amd64 0xF 0xD PREFETCH
n-i-bz  XML <unique> printing wrong
n-i-bz  Dirk r4359 (amd64 syscalls from trunk)
110591  amd64 and x86: rdtsc not implemented properly
n-i-bz  Nick r4384 (stub implementations of Addrcheck and Helgrind)
110652  AMD64 valgrind crashes on cwtd instruction
110653  AMD64 valgrind crashes on sarb $0x4,foo(%rip) instruction
110656  PATH=/usr/bin::/bin valgrind foobar stats ./fooba
110657  Small test fixes
110671  vex x86->IR: unhandled instruction bytes: 0xF3 0xC3 (rep ret)
n-i-bz  Nick (Cachegrind should not assert when it encounters a client
        request.)
110685  amd64->IR: unhandled instruction bytes: 0xE1 0x56 (loope Jb)
110830  configuring with --host fails to build 32 bit on 64 bit target
110875  Assertion when execve fails
n-i-bz  Updates to Memcheck manual
n-i-bz  Fixed broken malloc_usable_size()
110898  opteron instructions missing: btq btsq btrq bsfq
110954  x86->IR: unhandled instruction bytes: 0xE2 0xF6 (loop Jb)
n-i-bz  Make suppressions work for "???" lines in stacktraces.
111006  bogus warnings from linuxthreads
111092  x86: dis_Grp2(Reg): unhandled case(x86) 
111231  sctp_getladdrs() and sctp_getpaddrs() returns uninitialized
        memory
111102  (comment #4)   Fixed 64-bit unclean "silly arg" message
n-i-bz  vex x86->IR: unhandled instruction bytes: 0x14 0x0
n-i-bz  minor umount/fcntl wrapper fixes
111090  Internal Error running Massif
101204  noisy warning
111513  Illegal opcode for SSE instruction (x86 movups)
111555  VEX/Makefile: CC is set to gcc
n-i-bz  Fix XML bugs in FAQ

(3.0.1: 29 August 05,
        vex/branches/VEX_3_0_BRANCH r1367,
        valgrind/branches/VALGRIND_3_0_BRANCH r4574).



Release 3.0.0 (3 August 2005)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3.0.0 is a major overhaul of Valgrind.  The most significant user
visible change is that Valgrind now supports architectures other than
x86.  The new architectures it supports are AMD64 and PPC32, and the
infrastructure is present for other architectures to be added later.

AMD64 support works well, but has some shortcomings:

- It generally won't be as solid as the x86 version.  For example,
  support for more obscure instructions and system calls may be missing.
  We will fix these as they arise.

- Address space may be limited; see the point about
  position-independent executables below.

- If Valgrind is built on an AMD64 machine, it will only run 64-bit
  executables.  If you want to run 32-bit x86 executables under Valgrind
  on an AMD64, you will need to build Valgrind on an x86 machine and
  copy it to the AMD64 machine.  And it probably won't work if you do
  something tricky like exec'ing a 32-bit program from a 64-bit program
  while using --trace-children=yes.  We hope to improve this situation
  in the future.

The PPC32 support is very basic.  It may not work reliably even for
small programs, but it's a start.  Many thanks to Paul Mackerras for
his great work that enabled this support.  We are working to make
PPC32 usable as soon as possible.

Other user-visible changes:

- Valgrind is no longer built by default as a position-independent
  executable (PIE), as this caused too many problems.

  Without PIE enabled, AMD64 programs will only be able to access 2GB of
  address space.  We will fix this eventually, but not for the moment.
  
  Use --enable-pie at configure-time to turn this on.

- Support for programs that use stack-switching has been improved.  Use
  the --max-stackframe flag for simple cases, and the
  VALGRIND_STACK_REGISTER, VALGRIND_STACK_DEREGISTER and
  VALGRIND_STACK_CHANGE client requests for trickier cases.

- Support for programs that use self-modifying code has been improved,
  in particular programs that put temporary code fragments on the stack.
  This helps for C programs compiled with GCC that use nested functions,
  and also Ada programs.  This is controlled with the --smc-check
  flag, although the default setting should work in most cases.

- Output can now be printed in XML format.  This should make it easier
  for tools such as GUI front-ends and automated error-processing
  schemes to use Valgrind output as input.  The --xml flag controls this.
  As part of this change, ELF directory information is read from executables,
  so absolute source file paths are available if needed.

- Programs that allocate many heap blocks may run faster, due to
  improvements in certain data structures.

- Addrcheck is currently not working.  We hope to get it working again
  soon.  Helgrind is still not working, as was the case for the 2.4.0
  release.

- The JITter has been completely rewritten, and is now in a separate
  library, called Vex.  This enabled a lot of the user-visible changes,
  such as new architecture support.  The new JIT unfortunately translates
  more slowly than the old one, so programs may take longer to start.
  We believe the code quality is produces is about the same, so once
  started, programs should run at about the same speed.  Feedback about
  this would be useful.

  On the plus side, Vex and hence Memcheck tracks value flow properly
  through floating point and vector registers, something the 2.X line
  could not do.  That means that Memcheck is much more likely to be
  usably accurate on vectorised code.

- There is a subtle change to the way exiting of threaded programs
  is handled.  In 3.0, Valgrind's final diagnostic output (leak check,
  etc) is not printed until the last thread exits.  If the last thread
  to exit was not the original thread which started the program, any
  other process wait()-ing on this one to exit may conclude it has
  finished before the diagnostic output is printed.  This may not be
  what you expect.  2.X had a different scheme which avoided this
  problem, but caused deadlocks under obscure circumstances, so we
  are trying something different for 3.0.

- Small changes in control log file naming which make it easier to
  use valgrind for debugging MPI-based programs.  The relevant
  new flags are --log-file-exactly= and --log-file-qualifier=.

- As part of adding AMD64 support, DWARF2 CFI-based stack unwinding
  support was added.  In principle this means Valgrind can produce
  meaningful backtraces on x86 code compiled with -fomit-frame-pointer
  providing you also compile your code with -fasynchronous-unwind-tables.

- The documentation build system has been completely redone.
  The documentation masters are now in XML format, and from that
  HTML, PostScript and PDF documentation is generated.  As a result
  the manual is now available in book form.  Note that the
  documentation in the source tarballs is pre-built, so you don't need
  any XML processing tools to build Valgrind from a tarball.

Changes that are not user-visible:

- The code has been massively overhauled in order to modularise it.
  As a result we hope it is easier to navigate and understand.

- Lots of code has been rewritten.

BUGS FIXED:

110046  sz == 4 assertion failed 
109810  vex amd64->IR: unhandled instruction bytes: 0xA3 0x4C 0x70 0xD7
109802  Add a plausible_stack_size command-line parameter ?
109783  unhandled ioctl TIOCMGET (running hw detection tool discover) 
109780  unhandled ioctl BLKSSZGET (running fdisk -l /dev/hda)
109718  vex x86->IR: unhandled instruction: ffreep 
109429  AMD64 unhandled syscall: 127 (sigpending)
109401  false positive uninit in strchr from ld-linux.so.2
109385  "stabs" parse failure 
109378  amd64: unhandled instruction REP NOP
109376  amd64: unhandled instruction LOOP Jb 
109363  AMD64 unhandled instruction bytes 
109362  AMD64 unhandled syscall: 24 (sched_yield)
109358  fork() won't work with valgrind-3.0 SVN
109332  amd64 unhandled instruction: ADC Ev, Gv
109314  Bogus memcheck report on amd64
108883  Crash; vg_memory.c:905 (vgPlain_init_shadow_range):
        Assertion `vgPlain_defined_init_shadow_page()' failed.
108349  mincore syscall parameter checked incorrectly 
108059  build infrastructure: small update
107524  epoll_ctl event parameter checked on EPOLL_CTL_DEL
107123  Vex dies with unhandled instructions: 0xD9 0x31 0xF 0xAE
106841  auxmap & openGL problems
106713  SDL_Init causes valgrind to exit
106352  setcontext and makecontext not handled correctly 
106293  addresses beyond initial client stack allocation 
        not checked in VALGRIND_DO_LEAK_CHECK
106283  PIE client programs are loaded at address 0
105831  Assertion `vgPlain_defined_init_shadow_page()' failed.
105039  long run-times probably due to memory manager 
104797  valgrind needs to be aware of BLKGETSIZE64
103594  unhandled instruction: FICOM
103320  Valgrind 2.4.0 fails to compile with gcc 3.4.3 and -O0
103168  potentially memory leak in coregrind/ume.c 
102039  bad permissions for mapped region at address 0xB7C73680
101881  weird assertion problem
101543  Support fadvise64 syscalls
75247   x86_64/amd64 support (the biggest "bug" we have ever fixed)

(3.0RC1: 27 July   05, vex r1303, valgrind r4283).
(3.0.0:   3 August 05, vex r1313, valgrind r4316).



Stable release 2.4.0 (March 2005) -- CHANGES RELATIVE TO 2.2.0
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2.4.0 brings many significant changes and bug fixes.  The most
significant user-visible change is that we no longer supply our own
pthread implementation.  Instead, Valgrind is finally capable of
running the native thread library, either LinuxThreads or NPTL.

This means our libpthread has gone, along with the bugs associated
with it.  Valgrind now supports the kernel's threading syscalls, and
lets you use your standard system libpthread.  As a result:

* There are many fewer system dependencies and strange library-related
  bugs.  There is a small performance improvement, and a large
  stability improvement.

* On the downside, Valgrind can no longer report misuses of the POSIX
  PThreads API.  It also means that Helgrind currently does not work.
  We hope to fix these problems in a future release.

Note that running the native thread libraries does not mean Valgrind
is able to provide genuine concurrent execution on SMPs.  We still
impose the restriction that only one thread is running at any given
time.

There are many other significant changes too:

* Memcheck is (once again) the default tool.

* The default stack backtrace is now 12 call frames, rather than 4.

* Suppressions can have up to 25 call frame matches, rather than 4.

* Memcheck and Addrcheck use less memory.  Under some circumstances,
  they no longer allocate shadow memory if there are large regions of
  memory with the same A/V states - such as an mmaped file.

* The memory-leak detector in Memcheck and Addrcheck has been
  improved.  It now reports more types of memory leak, including
  leaked cycles.  When reporting leaked memory, it can distinguish
  between directly leaked memory (memory with no references), and
  indirectly leaked memory (memory only referred to by other leaked
  memory).

* Memcheck's confusion over the effect of mprotect() has been fixed:
  previously mprotect could erroneously mark undefined data as
  defined.

* Signal handling is much improved and should be very close to what
  you get when running natively.  

  One result of this is that Valgrind observes changes to sigcontexts
  passed to signal handlers.  Such modifications will take effect when
  the signal returns.  You will need to run with --single-step=yes to
  make this useful.

* Valgrind is built in Position Independent Executable (PIE) format if
  your toolchain supports it.  This allows it to take advantage of all
  the available address space on systems with 4Gbyte user address
  spaces.

* Valgrind can now run itself (requires PIE support).

* Syscall arguments are now checked for validity.  Previously all
  memory used by syscalls was checked, but now the actual values
  passed are also checked.

* Syscall wrappers are more robust against bad addresses being passed
  to syscalls: they will fail with EFAULT rather than killing Valgrind
  with SIGSEGV.

* Because clone() is directly supported, some non-pthread uses of it
  will work.  Partial sharing (where some resources are shared, and
  some are not) is not supported.

* open() and readlink() on /proc/self/exe are supported.

BUGS FIXED:

88520   pipe+fork+dup2 kills the main program
88604 	Valgrind Aborts when using $VALGRIND_OPTS and user progra...
88614 	valgrind: vg_libpthread.c:2323 (read): Assertion `read_pt...
88703 	Stabs parser fails to handle ";"
88886 	ioctl wrappers for TIOCMBIS and TIOCMBIC
89032 	valgrind pthread_cond_timedwait fails
89106 	the 'impossible' happened
89139 	Missing sched_setaffinity & sched_getaffinity
89198 	valgrind lacks support for SIOCSPGRP and SIOCGPGRP
89263 	Missing ioctl translations for scsi-generic and CD playing
89440 	tests/deadlock.c line endings
89481 	`impossible' happened: EXEC FAILED
89663 	valgrind 2.2.0 crash on Redhat 7.2
89792 	Report pthread_mutex_lock() deadlocks instead of returnin...
90111 	statvfs64 gives invalid error/warning
90128 	crash+memory fault with stabs generated by gnat for a run...
90778 	VALGRIND_CHECK_DEFINED() not as documented in memcheck.h
90834 	cachegrind crashes at end of program without reporting re...
91028 	valgrind: vg_memory.c:229 (vgPlain_unmap_range): Assertio...
91162 	valgrind crash while debugging drivel 1.2.1
91199 	Unimplemented function
91325 	Signal routing does not propagate the siginfo structure
91599 	Assertion `cv == ((void *)0)'
91604 	rw_lookup clears orig and sends the NULL value to rw_new
91821 	Small problems building valgrind with $top_builddir ne $t...
91844 	signal 11 (SIGSEGV) at get_tcb (libpthread.c:86) in corec...
92264 	UNIMPLEMENTED FUNCTION: pthread_condattr_setpshared
92331 	per-target flags necessitate AM_PROG_CC_C_O
92420 	valgrind doesn't compile with linux 2.6.8.1/9
92513 	Valgrind 2.2.0 generates some warning messages
92528 	vg_symtab2.c:170 (addLoc): Assertion `loc->size > 0' failed.
93096 	unhandled ioctl 0x4B3A and 0x5601
93117 	Tool and core interface versions do not match
93128 	Can't run valgrind --tool=memcheck because of unimplement...
93174 	Valgrind can crash if passed bad args to certain syscalls
93309 	Stack frame in new thread is badly aligned
93328 	Wrong types used with sys_sigprocmask()
93763 	/usr/include/asm/msr.h is missing
93776 	valgrind: vg_memory.c:508 (vgPlain_find_map_space): Asser...
93810 	fcntl() argument checking a bit too strict
94378 	Assertion `tst->sigqueue_head != tst->sigqueue_tail' failed.
94429 	valgrind 2.2.0 segfault with mmap64 in glibc 2.3.3
94645 	Impossible happened: PINSRW mem
94953 	valgrind: the `impossible' happened: SIGSEGV
95667 	Valgrind does not work with any KDE app
96243 	Assertion 'res==0' failed
96252 	stage2 loader of valgrind fails to allocate memory
96520 	All programs crashing at _dl_start (in /lib/ld-2.3.3.so) ...
96660 	ioctl CDROMREADTOCENTRY causes bogus warnings
96747 	After looping in a segfault handler, the impossible happens
96923 	Zero sized arrays crash valgrind trace back with SIGFPE
96948 	valgrind stops with assertion failure regarding mmap2
96966 	valgrind fails when application opens more than 16 sockets
97398 	valgrind: vg_libpthread.c:2667 Assertion failed
97407 	valgrind: vg_mylibc.c:1226 (vgPlain_safe_fd): Assertion `...
97427 	"Warning: invalid file descriptor -1 in syscall close()" ...
97785 	missing backtrace
97792 	build in obj dir fails - autoconf / makefile cleanup
97880 	pthread_mutex_lock fails from shared library (special ker...
97975 	program aborts without ang VG messages
98129 	Failed when open and close file 230000 times using stdio
98175 	Crashes when using valgrind-2.2.0 with a program using al...
98288 	Massif broken
98303 	UNIMPLEMENTED FUNCTION pthread_condattr_setpshared
98630 	failed--compilation missing warnings.pm, fails to make he...
98756 	Cannot valgrind signal-heavy kdrive X server
98966 	valgrinding the JVM fails with a sanity check assertion
99035 	Valgrind crashes while profiling
99142 	loops with message "Signal 11 being dropped from thread 0...
99195 	threaded apps crash on thread start (using QThread::start...
99348 	Assertion `vgPlain_lseek(core_fd, 0, 1) == phdrs[i].p_off...
99568 	False negative due to mishandling of mprotect
99738 	valgrind memcheck crashes on program that uses sigitimer
99923 	0-sized allocations are reported as leaks
99949 	program seg faults after exit()
100036 	"newSuperblock's request for 1048576 bytes failed"
100116 	valgrind: (pthread_cond_init): Assertion `sizeof(* cond) ...
100486 	memcheck reports "valgrind: the `impossible' happened: V...
100833 	second call to "mremap" fails with EINVAL
101156 	(vgPlain_find_map_space): Assertion `(addr & ((1 << 12)-1...
101173 	Assertion `recDepth >= 0 && recDepth < 500' failed
101291 	creating threads in a forked process fails
101313 	valgrind causes different behavior when resizing a window...
101423 	segfault for c++ array of floats
101562 	valgrind massif dies on SIGINT even with signal handler r...


Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.0.0
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2.2.0 brings nine months worth of improvements and bug fixes.  We
believe it to be a worthy successor to 2.0.0.  There are literally
hundreds of bug fixes and minor improvements.  There are also some
fairly major user-visible changes:

* A complete overhaul of handling of system calls and signals, and 
  their interaction with threads.  In general, the accuracy of the 
  system call, thread and signal simulations is much improved:

  - Blocking system calls behave exactly as they do when running
    natively (not on valgrind).  That is, if a syscall blocks only the
    calling thread when running natively, than it behaves the same on
    valgrind.  No more mysterious hangs because V doesn't know that some
    syscall or other, should block only the calling thread.

  - Interrupted syscalls should now give more faithful results.

  - Signal contexts in signal handlers are supported.

* Improvements to NPTL support to the extent that V now works 
  properly on NPTL-only setups.

* Greater isolation between Valgrind and the program being run, so
  the program is less likely to inadvertently kill Valgrind by
  doing wild writes.

* Massif: a new space profiling tool.  Try it!  It's cool, and it'll
  tell you in detail where and when your C/C++ code is allocating heap.
  Draws pretty .ps pictures of memory use against time.  A potentially
  powerful tool for making sense of your program's space use.

* File descriptor leakage checks.  When enabled, Valgrind will print out
  a list of open file descriptors on exit.

* Improved SSE2/SSE3 support.

* Time-stamped output; use --time-stamp=yes



Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.1.2
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2.2.0 is not much different from 2.1.2, released seven weeks ago.
A number of bugs have been fixed, most notably #85658, which gave
problems for quite a few people.  There have been many internal
cleanups, but those are not user visible.

The following bugs have been fixed since 2.1.2:

85658   Assert in coregrind/vg_libpthread.c:2326 (open64) !=
        (void*)0 failed
        This bug was reported multiple times, and so the following
        duplicates of it are also fixed: 87620, 85796, 85935, 86065, 
        86919, 86988, 87917, 88156

80716   Semaphore mapping bug caused by unmap (sem_destroy)
        (Was fixed prior to 2.1.2)

86987   semctl and shmctl syscalls family is not handled properly

86696   valgrind 2.1.2 + RH AS2.1 + librt

86730   valgrind locks up at end of run with assertion failure 
        in __pthread_unwind

86641   memcheck doesn't work with Mesa OpenGL/ATI on Suse 9.1
        (also fixes 74298, a duplicate of this)

85947   MMX/SSE unhandled instruction 'sfence'

84978   Wrong error "Conditional jump or move depends on
        uninitialised value" resulting from "sbbl %reg, %reg"

86254   ssort() fails when signed int return type from comparison is 
        too small to handle result of unsigned int subtraction

87089   memalign( 4, xxx) makes valgrind assert

86407   Add support for low-level parallel port driver ioctls.

70587   Add timestamps to Valgrind output? (wishlist)

84937   vg_libpthread.c:2505 (se_remap): Assertion `res == 0'
        (fixed prior to 2.1.2)

86317   cannot load libSDL-1.2.so.0 using valgrind

86989   memcpy from mac_replace_strmem.c complains about
        uninitialized pointers passed when length to copy is zero

85811   gnu pascal symbol causes segmentation fault; ok in 2.0.0

79138   writing to sbrk()'d memory causes segfault

77369   sched deadlock while signal received during pthread_join
        and the joined thread exited

88115   In signal handler for SIGFPE,  siginfo->si_addr is wrong 
        under Valgrind

78765   Massif crashes on app exit if FP exceptions are enabled

Additionally there are the following changes, which are not 
connected to any bug report numbers, AFAICS:

* Fix scary bug causing mis-identification of SSE stores vs
  loads and so causing memcheck to sometimes give nonsense results
  on SSE code.

* Add support for the POSIX message queue system calls.

* Fix to allow 32-bit Valgrind to run on AMD64 boxes.  Note: this does
  NOT allow Valgrind to work with 64-bit executables - only with 32-bit
  executables on an AMD64 box.

* At configure time, only check whether linux/mii.h can be processed 
  so that we don't generate ugly warnings by trying to compile it.

* Add support for POSIX clocks and timers.



Developer (cvs head) release 2.1.2 (18 July 2004)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2.1.2 contains four months worth of bug fixes and refinements.
Although officially a developer release, we believe it to be stable
enough for widespread day-to-day use.  2.1.2 is pretty good, so try it
first, although there is a chance it won't work.  If so then try 2.0.0
and tell us what went wrong."  2.1.2 fixes a lot of problems present
in 2.0.0 and is generally a much better product.

Relative to 2.1.1, a large number of minor problems with 2.1.1 have
been fixed, and so if you use 2.1.1 you should try 2.1.2.  Users of
the last stable release, 2.0.0, might also want to try this release.

The following bugs, and probably many more, have been fixed.  These
are listed at http://bugs.kde.org.  Reporting a bug for valgrind in
the http://bugs.kde.org is much more likely to get you a fix than
mailing developers directly, so please continue to keep sending bugs
there.

76869   Crashes when running any tool under Fedora Core 2 test1
        This fixes the problem with returning from a signal handler 
        when VDSOs are turned off in FC2.

69508   java 1.4.2 client fails with erroneous "stack size too small".
        This fix makes more of the pthread stack attribute related 
        functions work properly.  Java still doesn't work though.

71906   malloc alignment should be 8, not 4
        All memory returned by malloc/new etc is now at least
        8-byte aligned.

81970   vg_alloc_ThreadState: no free slots available
        (closed because the workaround is simple: increase
         VG_N_THREADS, rebuild and try again.)

78514   Conditional jump or move depends on uninitialized value(s)
        (a slight mishanding of FP code in memcheck)

77952   pThread Support (crash) (due to initialisation-ordering probs)
        (also 85118)

80942   Addrcheck wasn't doing overlap checking as it should.
78048   return NULL on malloc/new etc failure, instead of asserting
73655   operator new() override in user .so files often doesn't get picked up
83060   Valgrind does not handle native kernel AIO
69872   Create proper coredumps after fatal signals
82026   failure with new glibc versions: __libc_* functions are not exported
70344   UNIMPLEMENTED FUNCTION: tcdrain 
81297   Cancellation of pthread_cond_wait does not require mutex
82872   Using debug info from additional packages (wishlist)
83025   Support for ioctls FIGETBSZ and FIBMAP
83340   Support for ioctl HDIO_GET_IDENTITY
79714   Support for the semtimedop system call.
77022   Support for ioctls FBIOGET_VSCREENINFO and FBIOGET_FSCREENINFO
82098   hp2ps ansification (wishlist)
83573   Valgrind SIGSEGV on execve
82999   show which cmdline option was erroneous (wishlist)
83040   make valgrind VPATH and distcheck-clean (wishlist)
83998   Assertion `newfd > vgPlain_max_fd' failed (see below)
82722   Unchecked mmap in as_pad leads to mysterious failures later
78958   memcheck seg faults while running Mozilla 
85416   Arguments with colon (e.g. --logsocket) ignored


Additionally there are the following changes, which are not 
connected to any bug report numbers, AFAICS:

* Rearranged address space layout relative to 2.1.1, so that
  Valgrind/tools will run out of memory later than currently in many
  circumstances.  This is good news esp. for Calltree.  It should
  be possible for client programs to allocate over 800MB of
  memory when using memcheck now.

* Improved checking when laying out memory.  Should hopefully avoid
  the random segmentation faults that 2.1.1 sometimes caused.

* Support for Fedora Core 2 and SuSE 9.1.  Improvements to NPTL
  support to the extent that V now works properly on NPTL-only setups.

* Renamed the following options:
  --logfile-fd  -->  --log-fd
  --logfile     -->  --log-file
  --logsocket   -->  --log-socket
  to be consistent with each other and other options (esp. --input-fd).

* Add support for SIOCGMIIPHY, SIOCGMIIREG and SIOCSMIIREG ioctls and
  improve the checking of other interface related ioctls.

* Fix building with gcc-3.4.1.

* Remove limit on number of semaphores supported.

* Add support for syscalls: set_tid_address (258), acct (51).

* Support instruction "repne movs" -- not official but seems to occur.

* Implement an emulated soft limit for file descriptors in addition to
  the current reserved area, which effectively acts as a hard limit. The
  setrlimit system call now simply updates the emulated limits as best
  as possible - the hard limit is not allowed to move at all and just
  returns EPERM if you try and change it.  This should stop reductions
  in the soft limit causing assertions when valgrind tries to allocate
  descriptors from the reserved area.
  (This actually came from bug #83998).

* Major overhaul of Cachegrind implementation.  First user-visible change
  is that cachegrind.out files are now typically 90% smaller than they
  used to be;  code annotation times are correspondingly much smaller.
  Second user-visible change is that hit/miss counts for code that is
  unloaded at run-time is no longer dumped into a single "discard" pile,
  but accurately preserved.

* Client requests for telling valgrind about memory pools.



Developer (cvs head) release 2.1.1 (12 March 2004)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2.1.1 contains some internal structural changes needed for V's
long-term future.  These don't affect end-users.  Most notable
user-visible changes are:

* Greater isolation between Valgrind and the program being run, so
  the program is less likely to inadvertently kill Valgrind by
  doing wild writes.

* Massif: a new space profiling tool.  Try it!  It's cool, and it'll
  tell you in detail where and when your C/C++ code is allocating heap.
  Draws pretty .ps pictures of memory use against time.  A potentially
  powerful tool for making sense of your program's space use.

* Fixes for many bugs, including support for more SSE2/SSE3 instructions,
  various signal/syscall things, and various problems with debug
  info readers.

* Support for glibc-2.3.3 based systems.

We are now doing automatic overnight build-and-test runs on a variety
of distros.  As a result, we believe 2.1.1 builds and runs on:
Red Hat 7.2, 7.3, 8.0, 9, Fedora Core 1, SuSE 8.2, SuSE 9.


The following bugs, and probably many more, have been fixed.  These
are listed at http://bugs.kde.org.  Reporting a bug for valgrind in
the http://bugs.kde.org is much more likely to get you a fix than
mailing developers directly, so please continue to keep sending bugs
there.

69616   glibc 2.3.2 w/NPTL is massively different than what valgrind expects 
69856   I don't know how to instrument MMXish stuff (Helgrind)
73892   valgrind segfaults starting with Objective-C debug info 
        (fix for S-type stabs)
73145   Valgrind complains too much about close(<reserved fd>) 
73902   Shadow memory allocation seems to fail on RedHat 8.0 
68633   VG_N_SEMAPHORES too low (V itself was leaking semaphores)
75099   impossible to trace multiprocess programs 
76839   the `impossible' happened: disInstr: INT but not 0x80 ! 
76762   vg_to_ucode.c:3748 (dis_push_segreg): Assertion `sz == 4' failed. 
76747   cannot include valgrind.h in c++ program 
76223   parsing B(3,10) gave NULL type => impossible happens 
75604   shmdt handling problem 
76416   Problems with gcc 3.4 snap 20040225 
75614   using -gstabs when building your programs the `impossible' happened
75787   Patch for some CDROM ioctls CDORM_GET_MCN, CDROM_SEND_PACKET,
75294   gcc 3.4 snapshot's libstdc++ have unsupported instructions. 
        (REP RET)
73326   vg_symtab2.c:272 (addScopeRange): Assertion `range->size > 0' failed. 
72596   not recognizing __libc_malloc 
69489   Would like to attach ddd to running program 
72781   Cachegrind crashes with kde programs 
73055   Illegal operand at DXTCV11CompressBlockSSE2 (more SSE opcodes)
73026   Descriptor leak check reports port numbers wrongly 
71705   README_MISSING_SYSCALL_OR_IOCTL out of date 
72643   Improve support for SSE/SSE2 instructions 
72484   valgrind leaves it's own signal mask in place when execing 
72650   Signal Handling always seems to restart system calls 
72006   The mmap system call turns all errors in ENOMEM 
71781   gdb attach is pretty useless 
71180   unhandled instruction bytes: 0xF 0xAE 0x85 0xE8 
69886   writes to zero page cause valgrind to assert on exit 
71791   crash when valgrinding gimp 1.3 (stabs reader problem)
69783   unhandled syscall: 218 
69782   unhandled instruction bytes: 0x66 0xF 0x2B 0x80 
70385   valgrind fails if the soft file descriptor limit is less 
        than about 828
69529   "rep; nop" should do a yield 
70827   programs with lots of shared libraries report "mmap failed" 
        for some of them when reading symbols 
71028   glibc's strnlen is optimised enough to confuse valgrind 




Unstable (cvs head) release 2.1.0 (15 December 2003)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
For whatever it's worth, 2.1.0 actually seems pretty darn stable to me
(Julian).  It looks eminently usable, and given that it fixes some
significant bugs, may well be worth using on a day-to-day basis.
2.1.0 is known to build and pass regression tests on: SuSE 9, SuSE
8.2, RedHat 8.

2.1.0 most notably includes Jeremy Fitzhardinge's complete overhaul of
handling of system calls and signals, and their interaction with
threads.  In general, the accuracy of the system call, thread and
signal simulations is much improved.  Specifically:

- Blocking system calls behave exactly as they do when running
  natively (not on valgrind).  That is, if a syscall blocks only the
  calling thread when running natively, than it behaves the same on
  valgrind.  No more mysterious hangs because V doesn't know that some
  syscall or other, should block only the calling thread.

- Interrupted syscalls should now give more faithful results.

- Finally, signal contexts in signal handlers are supported.  As a
  result, konqueror on SuSE 9 no longer segfaults when notified of
  file changes in directories it is watching.

Other changes:

- Robert Walsh's file descriptor leakage checks.  When enabled,
  Valgrind will print out a list of open file descriptors on
  exit.  Along with each file descriptor, Valgrind prints out a stack
  backtrace of where the file was opened and any details relating to the
  file descriptor such as the file name or socket details.
  To use, give: --track-fds=yes

- Implemented a few more SSE/SSE2 instructions.

- Less crud on the stack when you do 'where' inside a GDB attach.

- Fixed the following bugs:
  68360: Valgrind does not compile against 2.6.0-testX kernels
  68525: CVS head doesn't compile on C90 compilers
  68566: pkgconfig support (wishlist)
  68588: Assertion `sz == 4' failed in vg_to_ucode.c (disInstr)
  69140: valgrind not able to explicitly specify a path to a binary. 
  69432: helgrind asserts encountering a MutexErr when there are 
         EraserErr suppressions

- Increase the max size of the translation cache from 200k average bbs
  to 300k average bbs.  Programs on the size of OOo (680m17) are
  thrashing the cache at the smaller size, creating large numbers of
  retranslations and wasting significant time as a result.



Stable release 2.0.0 (5 Nov 2003)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

2.0.0 improves SSE/SSE2 support, fixes some minor bugs, and
improves support for SuSE 9 and the Red Hat "Severn" beta.

- Further improvements to SSE/SSE2 support.  The entire test suite of
  the GNU Scientific Library (gsl-1.4) compiled with Intel Icc 7.1
  20030307Z '-g -O -xW' now works.  I think this gives pretty good
  coverage of SSE/SSE2 floating point instructions, or at least the
  subset emitted by Icc.

- Also added support for the following instructions:
    MOVNTDQ UCOMISD UNPCKLPS UNPCKHPS SQRTSS
    PUSH/POP %{FS,GS}, and PUSH %CS (Nb: there is no POP %CS).

- CFI support for GDB version 6.  Needed to enable newer GDBs
  to figure out where they are when using --gdb-attach=yes.

- Fix this:
      mc_translate.c:1091 (memcheck_instrument): Assertion
      `u_in->size == 4 || u_in->size == 16' failed.

- Return an error rather than panicing when given a bad socketcall.

- Fix checking of syscall rt_sigtimedwait().

- Implement __NR_clock_gettime (syscall 265).  Needed on Red Hat Severn.

- Fixed bug in overlap check in strncpy() -- it was assuming the src was 'n'
  bytes long, when it could be shorter, which could cause false
  positives.

- Support use of select() for very large numbers of file descriptors.

- Don't fail silently if the executable is statically linked, or is
  setuid/setgid. Print an error message instead.

- Support for old DWARF-1 format line number info.



Snapshot 20031012 (12 October 2003)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Three months worth of bug fixes, roughly.  Most significant single
change is improved SSE/SSE2 support, mostly thanks to Dirk Mueller.

20031012 builds on Red Hat Fedora ("Severn") but doesn't really work
(curiosly, mozilla runs OK, but a modest "ls -l" bombs).  I hope to
get a working version out soon.  It may or may not work ok on the
forthcoming SuSE 9; I hear positive noises about it but haven't been
able to verify this myself (not until I get hold of a copy of 9).

A detailed list of changes, in no particular order:

- Describe --gen-suppressions in the FAQ.

- Syscall __NR_waitpid supported.

- Minor MMX bug fix.

- -v prints program's argv[] at startup.

- More glibc-2.3 suppressions.

- Suppressions for stack underrun bug(s) in the c++ support library
  distributed with Intel Icc 7.0.

- Fix problems reading /proc/self/maps.

- Fix a couple of messages that should have been suppressed by -q, 
  but weren't.

- Make Addrcheck understand "Overlap" suppressions.

- At startup, check if program is statically linked and bail out if so.

- Cachegrind: Auto-detect Intel Pentium-M, also VIA Nehemiah

- Memcheck/addrcheck: minor speed optimisations

- Handle syscall __NR_brk more correctly than before.

- Fixed incorrect allocate/free mismatch errors when using
  operator new(unsigned, std::nothrow_t const&)
  operator new[](unsigned, std::nothrow_t const&)

- Support POSIX pthread spinlocks.

- Fixups for clean compilation with gcc-3.3.1.

- Implemented more opcodes: 
    - push %es
    - push %ds
    - pop %es
    - pop %ds
    - movntq
    - sfence
    - pshufw
    - pavgb
    - ucomiss
    - enter
    - mov imm32, %esp
    - all "in" and "out" opcodes
    - inc/dec %esp
    - A whole bunch of SSE/SSE2 instructions

- Memcheck: don't bomb on SSE/SSE2 code.


Snapshot 20030725 (25 July 2003)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Fixes some minor problems in 20030716.

- Fix bugs in overlap checking for strcpy/memcpy etc.

- Do overlap checking with Addrcheck as well as Memcheck.

- Fix this:
      Memcheck: the `impossible' happened:
      get_error_name: unexpected type

- Install headers needed to compile new skins.

- Remove leading spaces and colon in the LD_LIBRARY_PATH / LD_PRELOAD
  passed to non-traced children.

- Fix file descriptor leak in valgrind-listener.

- Fix longstanding bug in which the allocation point of a 
  block resized by realloc was not correctly set.  This may
  have caused confusing error messages.


Snapshot 20030716 (16 July 2003)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

20030716 is a snapshot of our current CVS head (development) branch.
This is the branch which will become valgrind-2.0.  It contains
significant enhancements over the 1.9.X branch.

Despite this being a snapshot of the CVS head, it is believed to be
quite stable -- at least as stable as 1.9.6 or 1.0.4, if not more so
-- and therefore suitable for widespread use.  Please let us know asap
if it causes problems for you.

Two reasons for releasing a snapshot now are:

- It's been a while since 1.9.6, and this snapshot fixes
  various problems that 1.9.6 has with threaded programs 
  on glibc-2.3.X based systems.

- So as to make available improvements in the 2.0 line.

Major changes in 20030716, as compared to 1.9.6:

- More fixes to threading support on glibc-2.3.1 and 2.3.2-based
  systems (SuSE 8.2, Red Hat 9).  If you have had problems
  with inconsistent/illogical behaviour of errno, h_errno or the DNS
  resolver functions in threaded programs, 20030716 should improve
  matters.  This snapshot seems stable enough to run OpenOffice.org
  1.1rc on Red Hat 7.3, SuSE 8.2 and Red Hat 9, and that's a big
  threaded app if ever I saw one.

- Automatic generation of suppression records; you no longer
  need to write them by hand.  Use --gen-suppressions=yes.

- strcpy/memcpy/etc check their arguments for overlaps, when
  running with the Memcheck or Addrcheck skins.

- malloc_usable_size() is now supported.

- new client requests:
    - VALGRIND_COUNT_ERRORS, VALGRIND_COUNT_LEAKS: 
      useful with regression testing
    - VALGRIND_NON_SIMD_CALL[0123]: for running arbitrary functions 
      on real CPU (use with caution!)

- The GDB attach mechanism is more flexible.  Allow the GDB to
  be run to be specified by --gdb-path=/path/to/gdb, and specify
  which file descriptor V will read its input from with
  --input-fd=<number>.

- Cachegrind gives more accurate results (wasn't tracking instructions in
  malloc() and friends previously, is now).

- Complete support for the MMX instruction set.

- Partial support for the SSE and SSE2 instruction sets.  Work for this
  is ongoing.  About half the SSE/SSE2 instructions are done, so
  some SSE based programs may work.  Currently you need to specify
  --skin=addrcheck.  Basically not suitable for real use yet.

- Significant speedups (10%-20%) for standard memory checking.

- Fix assertion failure in pthread_once().

- Fix this:
    valgrind: vg_intercept.c:598 (vgAllRoadsLeadToRome_select): 
              Assertion `ms_end >= ms_now' failed.

- Implement pthread_mutexattr_setpshared.

- Understand Pentium 4 branch hints.  Also implemented a couple more
  obscure x86 instructions.

- Lots of other minor bug fixes.

- We have a decent regression test system, for the first time.
  This doesn't help you directly, but it does make it a lot easier
  for us to track the quality of the system, especially across
  multiple linux distributions.  

  You can run the regression tests with 'make regtest' after 'make
  install' completes.  On SuSE 8.2 and Red Hat 9 I get this:
 
     == 84 tests, 0 stderr failures, 0 stdout failures ==

  On Red Hat 8, I get this:

     == 84 tests, 2 stderr failures, 1 stdout failure ==
     corecheck/tests/res_search               (stdout)
     memcheck/tests/sigaltstack               (stderr)

  sigaltstack is probably harmless.  res_search doesn't work
  on R H 8 even running natively, so I'm not too worried.   

  On Red Hat 7.3, a glibc-2.2.5 system, I get these harmless failures:

     == 84 tests, 2 stderr failures, 1 stdout failure ==
     corecheck/tests/pth_atfork1              (stdout)
     corecheck/tests/pth_atfork1              (stderr)
     memcheck/tests/sigaltstack               (stderr)

  You need to run on a PII system, at least, since some tests
  contain P6-specific instructions, and the test machine needs
  access to the internet so that corecheck/tests/res_search
  (a test that the DNS resolver works) can function.

As ever, thanks for the vast amount of feedback :) and bug reports :(
We may not answer all messages, but we do at least look at all of
them, and tend to fix the most frequently reported bugs.



Version 1.9.6 (7 May 2003 or thereabouts)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Major changes in 1.9.6:

- Improved threading support for glibc >= 2.3.2 (SuSE 8.2,
  RedHat 9, to name but two ...)  It turned out that 1.9.5
  had problems with threading support on glibc >= 2.3.2,
  usually manifested by threaded programs deadlocking in system calls,
  or running unbelievably slowly.  Hopefully these are fixed now.  1.9.6
  is the first valgrind which gives reasonable support for
  glibc-2.3.2.  Also fixed a 2.3.2 problem with pthread_atfork().

- Majorly expanded FAQ.txt.  We've added workarounds for all
  common problems for which a workaround is known.

Minor changes in 1.9.6:

- Fix identification of the main thread's stack.  Incorrect
  identification of it was causing some on-stack addresses to not get
  identified as such.  This only affected the usefulness of some error
  messages; the correctness of the checks made is unchanged.

- Support for kernels >= 2.5.68.

- Dummy implementations of __libc_current_sigrtmin, 
  __libc_current_sigrtmax and __libc_allocate_rtsig, hopefully
  good enough to keep alive programs which previously died for lack of
  them.

- Fix bug in the VALGRIND_DISCARD_TRANSLATIONS client request.

- Fix bug in the DWARF2 debug line info loader, when instructions 
  following each other have source lines far from each other 
  (e.g. with inlined functions).

- Debug info reading: read symbols from both "symtab" and "dynsym"
  sections, rather than merely from the one that comes last in the
  file.

- New syscall support: prctl(), creat(), lookup_dcookie().

- When checking calls to accept(), recvfrom(), getsocketopt(),
  don't complain if buffer values are NULL.

- Try and avoid assertion failures in
  mash_LD_PRELOAD_and_LD_LIBRARY_PATH.

- Minor bug fixes in cg_annotate.



Version 1.9.5 (7 April 2003)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~

It occurs to me that it would be helpful for valgrind users to record
in the source distribution the changes in each release.  So I now
attempt to mend my errant ways :-)  Changes in this and future releases
will be documented in the NEWS file in the source distribution.

Major changes in 1.9.5:

- (Critical bug fix): Fix a bug in the FPU simulation.  This was
  causing some floating point conditional tests not to work right.
  Several people reported this.  If you had floating point code which
  didn't work right on 1.9.1 to 1.9.4, it's worth trying 1.9.5.

- Partial support for Red Hat 9.  RH9 uses the new Native Posix 
  Threads Library (NPTL), instead of the older LinuxThreads.  
  This potentially causes problems with V which will take some
  time to correct.  In the meantime we have partially worked around
  this, and so 1.9.5 works on RH9.  Threaded programs still work,
  but they may deadlock, because some system calls (accept, read,
  write, etc) which should be nonblocking, in fact do block.  This
  is a known bug which we are looking into.

  If you can, your best bet (unfortunately) is to avoid using 
  1.9.5 on a Red Hat 9 system, or on any NPTL-based distribution.
  If your glibc is 2.3.1 or earlier, you're almost certainly OK.

Minor changes in 1.9.5:

- Added some #errors to valgrind.h to ensure people don't include
  it accidentally in their sources.  This is a change from 1.0.X
  which was never properly documented.  The right thing to include
  is now memcheck.h.  Some people reported problems and strange
  behaviour when (incorrectly) including valgrind.h in code with 
  1.9.1 -- 1.9.4.  This is no longer possible.

- Add some __extension__ bits and pieces so that gcc configured
  for valgrind-checking compiles even with -Werror.  If you
  don't understand this, ignore it.  Of interest to gcc developers
  only.

- Removed a pointless check which caused problems interworking 
  with Clearcase.  V would complain about shared objects whose
  names did not end ".so", and refuse to run.  This is now fixed.
  In fact it was fixed in 1.9.4 but not documented.

- Fixed a bug causing an assertion failure of "waiters == 1"
  somewhere in vg_scheduler.c, when running large threaded apps,
  notably MySQL.

- Add support for the munlock system call (124).

Some comments about future releases:

1.9.5 is, we hope, the most stable Valgrind so far.  It pretty much
supersedes the 1.0.X branch.  If you are a valgrind packager, please
consider making 1.9.5 available to your users.  You can regard the
1.0.X branch as obsolete: 1.9.5 is stable and vastly superior.  There
are no plans at all for further releases of the 1.0.X branch.

If you want a leading-edge valgrind, consider building the cvs head
(from SourceForge), or getting a snapshot of it.  Current cool stuff
going in includes MMX support (done); SSE/SSE2 support (in progress),
a significant (10-20%) performance improvement (done), and the usual
large collection of minor changes.  Hopefully we will be able to
improve our NPTL support, but no promises.