v0.88.1 31st October 2008:
-=-=-=-=-=-=-=-=-=-=-=-=-=
T (svnr) description
B (1080) TnFXSQLDBIPCMsgsI::ColumnData was being used inside a vector<> when
it has move only semantics
B (1081) QFileInfo::baseName() was returning one character too few. Thanks to CK
for reporting this.
B (1082) Copied fix from recent dlmalloc into nedmalloc - under high load bit
pinning was being corrupted on multiprocessor machines

+ (1080) Fixed implementation of move semantics for C++0x compilers. Added move
semantics to all container classes in TnFOX
* (1083) Initial set of fixes for Apple Mac OS X 10.5 support
* (1084) Full port to Mac OS X 10.5
* (1085) Merged FOX v1.6.34

v0.88 13th June 2008:
-=-=-=-=-=-=-=-=-=-=-
T (svnr) description
B (1041) fxendianswap4() wasn't reliable on GCC
B (1062) Unicode string rendering on Windows had magically broken itself again :(
B (1073) QFile and QMemMap were offsetting the file pointer by ludicrous
amounts when using CR/LF translation on 32 bit systems only

* (1036) Moved v1.4 branch into branches, replaced trunk with v1.6 branch
* (1037) Bumped TnFOX to v0.90
* (1038) Restored FXString::fill() which for some reason Jeroen removed
* (1039) Made QFile much more conservative when asking for permissions when
opening for read access
* (1041) Renamed all fxendianswapX() to fxendianswap() with typed parameter.
This avoids the compiler having to go to memory and thus makes far better
use of registers (which the x86 bswap instruction uses anyway)
* (1044) Bumped TnFOX back to v0.88 for a point release
* (1045) Improved build documentation
* (1046) Replaced Linux implementation of FXProcess::noOfProcessors() with
vastly improved version.
* (1046) Merged FOX v1.6.33
* (1047) Added MSVC9 project files for Windows libraries
* (1048) Merged SQLite v3.5.9
+ (1049) Added MSVC9 third party library binaries for convenience
* (1050) Added a few C++0x features (move constructors, static_assert,
__func__) for compilers which support them
* (1051) Updated build system to set parallel job numbers based on available
CPU cores
+ (1053) Added FXMEMALIGNED() for asking compiler to align variable data
+ (1054) Replaced Maths::Array with a SIMD optimised Maths::Vector plus
added new vector benchmark & test
+ (1055) Replaced Maths::FRandomness with a version written using the new
Maths::Vector code
+ (1059) Rewritten FXFSMonitor to use inotify on Linux instead of FAM
+ (1060) Improved UTF detection via testing for used & unused code planes
* (1065) Converted TnFOXTrans.txt into UTF-8 format and adjusted CppMunge.py
to always read & write in that format
+ (1067) Added automated tool for translating TnFOX translation files via
Google Translation
+ (1068) Replaced TnFOXTrans.txt with Google Translated version and added
that to core DLL

v0.87 11th July 2007:
-=-=-=-=-=-=-=-=-=-=-
B Fixed deadlock in FXFSMonitor
B FXINLINE_MUTEX_IMPLEMENTATION was broken
B Fixed critical section nedmalloc miscompile
B nedmalloc wasn't locking mspace if all mspaces were locked
B FXApp was failing on init when trying to list fonts when Xft wasn't
enabled
B Fixed daylight saving time problems in FXTime
B Fixed win32 stack backtracing in FXException which hadn't been working
reliably since we moved to MSVC7.1
B TnFXSQLDB_sqlite3 was adding one to the month of any dates read in
B FXFont wasn't degrading gracefully to alternative fonts if a font wasn't
available
B Due to changes in newer GCC's, fxassemblerops was not compiling i386
assembler implementations
B Fixed segfault in FXFSMonitor on process exit due to unremoved paths
B Moved FXException data into private structure to work around MSVC stack
over-allocation bug
B QSSLDevice wasn't 64 bit compatible and threw random errors :(
B FXFSMonitor was hanging on process exit on POSIX due to FAM/gamin
B FXNetwork::dnsReverseLookup was failing on Apple MacOS X
B Thread cancellation wasn't invoking cleanup handlers on Linux

* Removed many unnecessary fetches of per-thread data (slow on POSIX) in
FXException
* Made FXProcess::mappedFiles() cached
* Fixed up memory locking code to lock per VirtualAlloc region
* Disabled signals to threads (without it, gprof won't run)
* Merged FOX v1.6.4
+ Added kqueue implementation to FXFSMonitor
+ Ported TnFOX to Apple MacOS X (Intel only)
* Merged FOX v1.6.6
* Substantially improved python bindings
+ Made FXAtomicInt, and therefore TnFOX, portable to all GCC supported
architectures
+ Added modular building of TnFOX
* Renamed FXSQLDB* to TnFXSQLDB*
+ Updated build system so it will compile on generic architectures
* Merged FOX v1.6.9
+ Added Pol::destructiveCopyNoDelete and Pol::noCopyNoDelete
+ Added support and use of restrict keyword
* Fixed up linking tests against a static TnFOX library
+ Added FXGLColor and replaced use of FXVec4f where appropriate
+ Added FXGLVertices
+ Added Generic::SmallestValue. Fixed up Generic::BiggestValue for MSVC
+ Added Secure::FRandomness
+ Added Maths namespace and moved Secure::FRandomness into it. Added
functions and classes Array, Array2, normalrand, normaldist, mean,
variance, stddev, distribution to Maths namespace
+ Added 64 bit overloads for fxbitscan() and fxbitscanrev()
+ Added Maths::InfinityValue and Maths::NaNValue
* Merged FOX v1.6.18
* Merged FOX v1.6.25
* Fixed problem with newer scons (> v0.96.95) which prevented correct
detection of build environment
* Updated doxygen config to v1.5.2
+ Added incorporation of current SVN revision into source
* Added FXProcess::buildInfo() and broke out hostOS() and
hostOSDescription()
+ Added QChildProcess
* Changed semantics of QIODeviceS::waitForData() as obvious usage was
causing data corruption (the null termination on signalled)
* Integrated sqlite v3.3.17
+ Added TnFXSQLDBTransaction
* Fixed a build problem on Unix with newer versions of scons
+ Added stack backtracing for GCC builds to FXException
+ Added errorIfNotFound parameter to fxdemanglesymbol()
* Made all POSIX file descriptors allocated by TnFOX FD_CLOEXEC so child
processes won't inherit them. This duplicates the Win32 behaviour.
+ Added automated test suite
+ Added Windows Vista support
* Merged FOX v1.6.28
* Updated translations

v0.86 3rd April 2006
-=-=-=-=-=-=-=-=-=-=
B FXMemMap was unintentially broken on Linux for shared memory usage
B FXMemMap wasn't detecting mmap() failures at all on POSIX!
B FXMemoryPool contained a use of static data after static data deinit
B FXMtxHold wasn't being inlined on Win32
B Various FXException macros were using "e" as their temporary exception
which could cause conflict with local definitions of "e"
B FXIPCChannel wasn't retrying message sends if they part completed
R *** Issued snapshot 1
B FXPipe::writeBlock() wasn't detecting write errors on win32
B Config failed due to lack of build directory on virgin builds
B FXACL would hang on win32 if program was not running as administrator
B Fixed segfault in FXSSLDevice on process exit
B If missing zlib was causing compile errors
B Fixed occasional segfault in ~FXMemoryPool
R *** Issued snapshot 2
B i/o classes weren't resetting their acl on close() which caused
problems if instance was reused
B FXMemMap::setName() wasn't setting name of underlying FXFile
B FXSSLDevice was corrupting data if data length wasn't multiple of key size
B FXExceptionDialog would cancel instead of quit if Escape were pressed
during a fatal exception
B FXACLEntity static methods weren't threadsafe :(
B FXSSLPKey operator < and > wasn't comparing key contents
B FXSSLKey operator < and > wasn't comparing key contents
B On Win32, multiple read/write access wasn't being permitted
B FXACLIterator::atEnd() was missing
B FXACL thought Win32 null DACLs meant no access instead of total access
B FXFSMonitor was failing to invoke change handler if handler invocation
was already in progress
B FXMemMap was extending zero size files to one byte
B FXACL was losing the error code on win32
B FXProcess::getNsCount() was misrepresenting microseconds as nanoseconds
on POSIX
B FXACLEntity::group() of owner() and everything() wasn't correct on POSIX
R *** Issued snapshot 3
B malloc_dbg on MSVC wasn't detecting failure to allocate memory (segfault)
B Finally fixed QDictBase attempting to delete void * after all these years!
B FXString::arg() wasn't printing negative numbers
B QHostAddress stream operators didn't have FXAPI
B FXIPCChannel wasn't detecting messages sent in big-endian
B FXIPCChannel::doReception() was mis-reading messages under high load
B FXProcess::dllLoad was reiniting FXProcess static inits even when DLL was
already loaded
B QTransString segfaulted on serialisation if empty
B FXACLEntity::homeDirectory() was failing on win32
B QSSLDevice wasn't closing underlying device if open() failed
B QPipe wasn't deleting named pipes on close
B Fixed fatal error on process load when running under Terminal Services
B Fixed compile error in Boost.Python patches with throw() declarations
B QFileInfo wasn't detecting if a file got deleted during metadata read
B FXFSMonitor wasn't detecting if files vanished between directory scan
and metadata scan
B QMemMap wasn't opening existing shared memory with read ACL access
B Fixed deadlock when creating thread during thread creation upcall
R *** Issued snapshot 4
B Fixed some compilation errors in QPtrVector
R *** Issued snapshot 5
B Fixed assertion failure in FXDebugTarget
B QBlkSocket wasn't reporting read nor write errors on win32

+ Added ability to set external array directly in QMemArray and QByteArray
+ Added own definition of struct timeval to the FX namespace as had been
having consistent problems with it sometimes being defined on Win32 and
other times not :(
* Force inlined FXMtxHold on compilers that support such things
+ Made it possible for FXAtomicInt and FXMutex to be inlineable to all
source files
* Removed locking in various places where it wasn't needed
+ Added FXProcess::getNsCount()
+ Added operator!= to FXHostAddress
- Removed the FXHostAddress::isLoopback() method
+ Added the FXHostAddress::isLocalMachine() method
* Changed FXFont to always specify anti-aliased and cleartype on Win32
* Now replaces system allocator on FreeBSD too
* FXFSMonitor now can disable itself if it detects a broken FAM implementation
R *** Issued snapshot 1
+ Added FXACLEntity::homeDirectory()
+ Added FXDEPRECATED macro
+ Added support for MSVC8.0
+ Added FX::Pol::itRevCompare
* FXSSLKey::setAsymmetricKey() now returns *this
+ Added MSVC project files for libjpeg, libtiff, zlib and placeholder
directories for openssl and libpng
- Removed administrator-only code in FXACL
+ Added Windows x64 capability to build system and fixed loads of 64 bit
warnings
* Modified QPtrVectorIterator to not inherit vector::iterator
* Reimplemented QPtrList::sort() with QValueListQSort (as list::sort()) is
broken on MSVC8)
* Reimplemented Windows version of FXAtomicInt in terms of Interlocked*
intrinsics (as inline assembler not supported on x64). Gives a 17% performance
increase in Tn TestIO due to skipping register dumping around _asm.
- Removed long and unsigned long overloads as they were unnecessary
* Fixed some small x64 FreeBSD incompatibilities
R *** Issued snapshot 2
+ Added fxprefetchmem() inline functions
* Replaced many macros in fxdefs.h with inline functions so pyste can see them
+ Finally implemented AllFromHeader() in the Python bindings and added policies
for all remaining classes. The bindings are now complete, if untested.
* Renamed all vector friend functions so they are prefixed with 'vec'
* Added a series of FN postfixes to the error handling macros to take a
filename so that it can be reported in the error message. Altered various
i/o classes to support specifying the filename
* Added full namespace specifiers to the error handling macros
* Added FXBUTTON_DANGEROUS to FXButton
+ Added class FXHandedMsgBox
* Changed FXThreadPool to return handles instead of BoundFunctorV's lest
external code go delete them without telling FXThreadPool (a problem somewhere
inside Tn)
* Added support for seeking, mixed reads & writes and truncating to FXSSLDevice
+ Added QSortedList::merge()
+ Added ability to send null pointer to FXMtxHold and added UnlockAndRelock for
FXRWMutex
* Allowed const FXRWMutex to be taken by FXMtxHold
+ Added Generic::lockedAccessor
* Changed FXTime to always be FXlong on all platforms
* Changed FXFile::stampCreationMetadata() to writeMetadata()
+ Added QPtrList::replaceAtIter() and QPtrVector::replaceAtIter() for iterators
+ Added permissions() and setPermissions() to FXSSLDevice
+ Added reloadSize() to FXFile and FXMemMap
+ Added fileHeaderLen() to FXSSLDevice
* Marked loads of extra stuff as FXDEPRECATEDEXT ie; deprecated for external use
* Made FXRefedObject more efficient by avoiding a lock when referrers are unknown
+ Added FXRefedObject_DisableCount
* Replaced all Qt-compatible FX things with Q in order to make way for
integrating future versions of FOX (this is REVISION 457 onwards in SVN)
* Removed wide character POSIX code from FXFile and FXPipe. It was never used.
* Removed pseudo-unicode support from FXString and fxdefs.h
+ Added insert pos cache to FXString so arg() inserts are much quicker
+ Added FXUnicodify and applied it to all TnFOX extension classes. This makes
all extension classes ready for Unicode support when FOX v1.6 is merged.
* Added new class FXTime for more accurate times
+ Added FXFile::readMetadata() and reimplemented all other FXFile metadata
functions to use it. Reimplemented QFileInfo to use it.
R *** Issued snapshot 3
+ Added a copy of sqlite (http://www.sqlite.org/)
+ Added C API functions tnfxmalloc() etc. to FXMemoryPool
+ Added fxbitscan() and fxbitscanrev() with x86/x64 optimised versions
+ Added fxassemblerops.h and fxbitscan(), fxendianswap() etc to it
+ Added magic disable idx to Generic::dynamicAt
+ Added noDeleteExtArray to QMemArray
+ Added Generic::BiggestValue
+ Added FXSQLDB and FXSQLDBRegistry
+ Added FXSQLDB_sqlite3 driver
+ Added QPtrVector::extend()
+ Added Generic::hasSerialise and hasDeserialise
* Made all FXStream integral operators inline
* Made FXString serialisation inline
+ Added FXIPCChannelIndirector
+ Added FXSQLDB_ipc and FXSQLDBServer
* Made QMemArray serialisation routines optimise for POD contents
+ Added QBuffer::isNull()
* Made FXIPCChannel::sendMsg() allocate send buffers based on previous send
message size
+ Added FXRefedObject::resetDying()
* Made FXProcess::dllHandle easier to use
+ Added operator!() to Generic::typeInfoBase
+ Added new option "disableGUI" to build system with small mods to sources
to remove FXApp and FXObject dependencies
+ Added QTransString::empty()
+ Added FXRectangle operators < and >
+ Added Generic::IntegralLists::biggestUnsignedInt, biggestSignedInt,
biggestFloat and smallestUnsignedInt, smallestSignedInt, smallestFloat
* Disabled XIM if XTHREADS defined. No choice really. Word is that XIM
will never be fixed ever
* FXApp now chooses 8pt Tahoma, 8pt Arial, 8pt Luxi Sans or 8pt Helvetica
to better approximate Windows layouts
+ Added FXACL::resetPath()
* Altered FXACLEntity::asString()
* FXFile metadata functions now report link metadata rather than what the
link points to
+ Added FXProcess::mountablePartitions, mountPartition and unmountPartition
+ Added FXWinShellLink and FXWinJunctionPoint
* Fixed up some mistakes in the build system
* Added support for Boost v1.33
* Fixed up the python bindings
* Added support for OpenSSL v0.9.8 and made it statically linked in
* Merged SQLite3 v3.2.7
* Made changes to support MSVC8.0 RTM
+ Added FXApp::addDestructionUpcall() and FXApp::removeDestructionUpcall()
+ Added ability to use multiple processors when building
* Changed QTL constructors to be explicit. This traps incorrect usage of
iterator operators.
* Replaced ptmalloc2 with nedmalloc
R *** Issued snapshot 4
* Now enables thread caching win32 allocator variant when available
* Merged FOX v1.4.29
* Merged FOX v1.6.0 RC3. Extensive changes have been made.
R *** Issued snapshot 5
+ Added fastbuffer option to QBuffer
+ Added QThread::processorAffinity() and setProcessorAffinity()
* Added some ambiguity busters for when including fx.h and misc headers.
Changed None to NoEncryption for QSSLKey and QSSLPKey
+ Added the FOX test suite provided kindly by jayn <at> io <dot> com
* Added GCC 4.1 compatibility, specifically a number of aliasing constructs
were made safe and warnings fixed
+ Added UTF-16 and UTF-32 to UTF-8 text translation by reworking
QIODevice::applyCRLF() and QIODevice::removeCRLF() and all i/o classes
using them via IO_Translate
+ Added auto-UTF detection code to QIODevice::determineUnicodeType() and
added unicode type state to all QIODevice's. i/o devices now examine text
data and set their unicode type on open
+ Added FXIO::TextTranslation to mirror IO_Translate and added to Adie
* Updated to doxygen 1.4.6
* Integrated sqlite v3.3.4
* Updated to boost 1.33.1
+ Added UDP support to QBlkSocket plus a UDP test to TestDeviceIO
+ Added script for installing TnFOX docs into Visual Studio
+ Added noGUI support to python bindings (very useful for debugging)
+ Replaced python bindings generator (pyste) with pyplusplus. Thanks to Roman
Yakovenko for doing most of the work!
+ Added QMemArray::swap()
+ Added class QBZip2Device
* Updated embedded translation file
* Merged FOX v1.6.1
* Added FXERRH_ISFOXEXCEPTION and made all FOX exceptions print to stderr
and call abort()


v0.85 13th Feb 2005:
-=-=-=-=-=-=-=-=-=-=
B Fixed compile error in FXRefedObject.h when FXRefedObject subclass defines
its own noMoreReferrers().
B FXIPCChannel should have thrown away ack messages when sender did not
specifiy ack location
B FXPipe was redefining FXERRHWIN for itself and thus losing the changes
implemented in FXException::int_throwWinError()
B FXMemMap::mappedRegion() was failing when offset was in last mapped
region
B Anything in FXDir or FXFile taking an "acceptAbs" parameter wasn't
accepting absolute paths even when this parameter was true
B Failure to truncate a FXFile or FXMemMap didn't generate an error code
on Win32
B FXMemMap::mapIn() would map a zero extent if the file was empty.
B If an FXException was being rethrown, stacklevel wasn't being reset on
transference up a stack level and thus causing a segfault on that exception's
destruction
B Fixed some bad assembler in ptmalloc2 and FXThread.cxx which GCC 3.4.x was
complaining about
B FXThread wasn't setting up default nested exception handling frame causing
fatal app exit
B Fixed memory leak in FXTextCodecs inherited from FOX
B Fixed occasional segfault in FXFSMonitor due to insufficient data locking
B Fixed failure to maintain wait counts in FXWaitCondition on Win32
B Fixed FXBlkSocket not being cancellable on FreeBSD
B FXMemMap wasn't using /tmp before its memory names which caused breakage
on FreeBSD
B Finally fixed problem with FXIPCChannel cleanup causing early thread exit
on POSIX
B FXThread::id() wasn't returning correct value on FreeBSD
B FXPipe::maxAtomicLength() was incorrect on FreeBSD
B Finally fixed FXPipe::readBlock() causing segfault on Linux 2.6 kernels
when library was being used dynamically
B Fixed fatal exception in FXFSMonitor on POSIX when shut down
B Fixed "no registered converter to void *" error from BPL on thread exit
B Fixed no top-level FXERRH_TRY fatal error when exception thrown inside python
B Fixed python threads not being able to find TnFOX import library

* Made FXMemMap::mappedRegion() return current and following region
+ Added FXProcess::virtualAddrSpaceLeft()
* Added a SMP build option which defines FX_SMPBUILD. Altered FXAtomicInt
to use non locked instructions when not defined.
+ Added fastinc() and fastdec() to FXAtomicInt. Made FXMutex use them instead.
+ Made build system define FX_X86PROCESSOR to be processor revision wanted.
* Modified FXMutex to use exchange instead of compare exchange (faster). Now
FXMutex lock/unlock is 18.3% faster when contended and 3.3% faster uncontended.
* Modified QDictBase::resize() to use vector::swap instead of a deep copy.
* Added QDictBase::safeResize() and modified QDICTDYNRESIZE() to use it
* Modified FXRefedObject and FXRefingObject to not allow opening of new
references after destruction due to zero reference count has begun.
+ Added CPPMUNGE_NOEXTRACTERRORCODES to CppMunge.py
+ Added removeRef() and takeRef() to QSortedList
+ Added equality and comparison operators to QPtrListIterator and
QVectorListIterator (and thus QSortedListIterator)
* More or less rewrote nested exception framework runtime support to be both
more efficient and handle multiple nested exception throws simultaneously
+ Added QPtrList::sort() with user supplied predicate function
+ Added QDictBase::append() and operator+=
* Modified all code using HAVE_MOVECONSTRUCTORS to disable copy constructor
+ Added QValueList::append()
* Modified FXTransString to have translate() taking a language id
* Changed FXProcess::dllpath() to be dllPath()
+ Added push_back() and append() to QMemArray
+ Added operator=() to FXTransString
+ Added isUnique() and setUnique() to FXMemMap
+ Added the most useful setMutex() to FXRollback
- Removed FXDir's fast constructor as the nearly static functions aren't for
a reason :)
+ Added FXException::setConstructionBreak()
+ Added FXFile::join()
+ Added support for FreeBSD v5.3
+ Added FXFile::stampCreationMetadata()
+ Added #define FXREFINGOBJECT_DEBUGKNOWALLOC
+ Added support for 64 bit x86 architectures
+ Added failonfree() and unfailonfree()
- Removed over allocation feature from FXMemMap
+ Added FXMemMap::mappableSize() and FXMemMap::maximiseMappableSize()
* Merged FOX v1.3.20 and updated docs
+ Added ability to set PYTHON_INCLUDE and PYTHON_LIB in config.py
* Made FXProcess use new fast FXFile instantiations
* FXThread::id() and myId() now return 64 bit quantities as it was not
possible to fit a thread id on FreeBSD into a 32 bit container (due to
process local threads not being a kernel process)
* Improved error reporting in FXFSMonitor
* Moved FXSecure to use /dev/urandom instead of /dev/random on POSIX
as the latter never had enough data on Linux 2.6 kernels
* Updated python bindings for latest TnFOX
* Updated to use Boost v1.32
* Modified FXScrollBar to use 64 bit ranges - as suggested by jayn on
tnfox-discussion
* Merged FOX v1.3.26
* Merged FOX v1.4.0
* Merged FOX v1.4.1
* Changed FXThread assembly to use GCC local labels
* Updated embedded translation file

v0.80 12th July 2004:
-=-=-=-=-=-=-=-=-=-=-
B FXPython.h wasn't observing the -fvisibility GCC extension
B QDictBase was never deleting keys, so key list was always growing
and performance dropping with time
B QDictBaseIterator became invalidated with any removals from
QDictBase
B FXIPCChannel::setRegistry() was corrupting memory
B Fixed malcode in QDictBase::operator>>
B Fixed QIntDict and QPtrDict saving incompatible file formats when between
32 & 64 bit architectures
B Fixed malcode in QSortedList operator >>
B FXProcess::dllLoad wasn't trying default extensions on Windows
B FXIPCChannel::doReception lost any messages split across a pipe buffer
boundary
B FXFSMonitor was being overzealous in verifying a path exists on Windows
B FXFSMonitor wasn't error checking path add on Windows
B FXRefedObject::noMoreReferrers() was being recursed into
B Errors and Unhandled messages from remote side of FXIPCChannel were being
ignored
B FXThreadPool::cancel() wasn't cancelling timed dispatches
B FXBlkSocket when closing wasn't causing remote end to fault (SO_DONTLINGER
isn't much use :( )
B FXBlkSocket wasn't detecting graceful closure on Windows (and was looping
forever)
B FXThreadPool::cancel() was returning if functor was running causing it to
get deleted during execution
B TnFXApp::exit() would hang if no event loops had ever been created
B FXThreadPool::cancel() was deleting functor if it was a timed dispatch and
was yet to fire
B FXThreadPool::reset() was returning if functor was running causing it to
spuriously create another functor
B FXThreadPool::remove() wasn't causing recalculation of wait list on Windows
B Generic::TL::instance<> would cause the compiler to throw a "requested base
class is ambiguous" error during compilation if there were more than one
of the same type in a typelist instantiated using Generic::TL::instantiateH.
B FXMemMap was always deleting shared memory names on first close
B FXBlkSocket::close() would throw exception if connection had been lost
B Kernel handle leak in ~FXThreadPrivate() fixed
B FXMemoryPool was leaking entire 1Mb segments :(
B QSortedList::take() wasn't using binary search
B FXThreadPool::reset() wasn't moving entry to correct place in timed order
list
B "scons tests" wasn't working (thanks to Taras for reporting this)
B FXThreadPool::cancel() was deleting functor
B Generic::Functor's void * constructor unfortunately consumed
Generic::Functor::Tester *, causing a segfault.
B FXThreadPool::cancel() would delete wrong functor if activation time was
identical to another.
B TnFXApp::exit() wasn't quite working correctly
B Stopped FXIPCChannel from sending disconnect message if its registry is dead
B Fixed segfault when FXThread_DTHold was used in a self-destructing thread
B FXIPCChannel wasn't calling constructor for messages with type<=4 which carry
an id (eg; ErrorOccurred) - this led to segfaults
B When FXIPCChannel returned an ErrorOccurred, it also returned an Unhandled.
B TnFXApp's event loop subclasses were calling X11 even if it wasn't initialised
B FXPipe wouldn't work with FXIODeviceS::waitForData() on POSIX
B FXFSMonitor wasn't handling FAM being out of date with respect to what's being
monitored (now it ignores FAM).
B FXFSMonitor::remove() wasn't cancelling pending callbacks
B FXIPCChannel wasn't dynamically adjusting its msg ack lists
B FXPipe was losing data during first read after open
B QSortedList::findIter() wasn't returning null iterator if not found
B Fixed segfault in FXProcess::memoryLoad()
B Fixed memory leak in FXException
B QPtrListIterator::toLast() and QPtrVectorIterator::toLast() failed when list
was empty.
B QPtrVectorIterator wouldn't die when run off front of list
B Fixed major bug where direct use of a FXMemoryPool was not locking pool lock
B Fixed data corruption when multiple threads wrote to a FXSSLDevice concurrently
B Fixed segfault when POSIX release build python bindings DLL was loaded
B Fixed creating a new python interpreter causing a segfault
B FXWaitCondition wasn't adjusting wait count when wait timed out
B FXProcess::exepath() wasn't working on Linux 2.6 kernels

+ Added FXMemMap::mappedRegion()
* Made FXACLEntity default constructor public
+ Added FXSSLPKey::publicKeyFromString()
* Made FXSSLPKey and FXSSLKey have default constructors
+ Added Generic::TextToType
+ Added if(sp) and if(!sp) to Generic::Functor
+ Added Generic::findParent and findParentC
+ Added default constructors to all QDictBase subclasses
+ Added default constructors to iterators for QPtrList, QPtrVector and
QSortedList
+ Added routing field to FXIPCMsg
+ Added QDICTDYNRESIZE(dict) and fx2powerprimes()
+ Added ability for FXIPCChannel::doHandled() to send back exception
* Tighted up access specifiers for FXIPCMsg. Caught a few sliced destructions.
* Upgraded doxygen to v1.3.6 and fixed various doc errors
+ Added default constructor and FXStream serialisation to Generic::typeInfoBase
+ Added QInt64Dict
* Made major structural changes to the build system due to Tn requirements
+ Added FXIPCChannel::maxMsgSize(), setMaxMsgSize(), garbageMessageCount() and
setGarbageMessageCount()
* Stopped stupid "media insert" dialog appearing when FXACL queries a
removable media drive
* Made FXThreadPool destructor clean itself from timer master lists in case
not all dispatches were freed before destruction.
+ Added TnFXAppEventLoop::getEventLoop()
* Finalised support for the new GCC visibility infrastructure esp. FXDLLLOCAL
+ Added FXThread::primaryThread()
* Altered FXFSMonitor to pass before and after file info's (more useful).
* Added assert.h to FXGenericTools.h and asserting indirections through Generic::ptr
* Reimplemented Generic::instantiateH with a superior implementation lacking the
"ambiguous base class" error above but also results in noticeably faster compile
times.
* Split Generic::TraitsBasic from Generic::Traits
+ Added i/o device flag IO_DontUnlink and support to FXPipe and FXMemMap
+ Added facility for always building static libraries on Win32
+ TnFXApp now can dispatch messages to the main loop on POSIX just Win32. Both
systems now behave identically.
+ Added QPtrList::takeByIter, QPtrVector::takeByIter and did better forwarding
of QPtrList functions through QSortedList
* Slightly refined build process (thanks to Taras for suggesting improvements)
+ Added a console control handler for all TnFOX processes to FXProcess for Win32
+ Added int_referrers() to FXRefedObject so code can read the referrers list
+ Added loads of extra overloads to FXString::arg() and FXString::number()
+ Added visibility local to private classes for improved GCC code quality
* FXProcess::hostOSMemoryLoad() had been broken by Linux kernel 2.6.x
* Collapsed FXIPCMessagingBase back into FXIPCChannel
+ Added FXIPCChannel::invokeMsgHandler() to outsource message handling to a thread
pool.
+ Added FXIPCChannel::unknownMsgReceived(), doAsyncHandled(), restampMsgAndSend(),
lonelyMsgAckReceived() and makeUniqueMsgId().
+ Added support for IPC message tunnelling and per-msg endianess
* All FXGenericTools.h and FXRollback.h stuff taking variable parameters can now
take up to four parameters instead of three
* FXRollback.h stuff now does a no op if code pointer is null
+ Added QPtrListIterator::makeDead()
* Merged TMemPool into FXMemoryPool. Added global dynamic memory allocator overrides
for all TnFOX.
* Improved make support for Whole Program Optimisation (WPO).
+ Added an optional initial size parameter to FXBuffer
* FXMutex now uses a cache for kernel wait objects to vastly improve creation and
deletion speed. Also now spin counts with processor yield on uniprocessor machines.
* Enabled aliasing on MSVC which yields 5-10% faster code.
* Ran all Tn and TnFOX code through lint - fixed a number of small issues.
* Merged FOX v1.2.4
* TnFOX-ised FXString.
* Modified build system to test GCC for -fvisibility and enable if supported
* Updated embedded translation file

v0.75 9th April 2004:
-=-=-=-=-=-=-=-=-=-=-
Thanks to Mykhaylo Sorochan for reporting problems with FreeBSD.

B Malcode in QMemArray::sort fixed
B Malcode in FXRefedObject::PtrRef() fixed
B Malcode in Generic::ptr when comparing against zero fixed
B scons was using GCC v3.4 options incompatible with GCC v3.2/3.3
B SOL_TCP didn't exist on BSD so replaced with getprotobyname()
B MSG_NOSIGNAL is Linux only, so moved FXPipe signal catcher into
common code and integrated into FXBlkSocket
B FXACL::check() was failing on Win32 when the calling thread was not
the main process thread
B Fixed segfault caused by FXIPCMsgRegistry getting deleted before its
registrants
B FXACLEntity::isGroup() was totally broken on Win32
B FXACLEntity::root() on Win32 returned the root group as the user
B FXACLEntity::lookupUser() was totally broken on Win32
B Removed the include zutil.h in FXGZipDevice as BSD doesn't come with that
B pthreads and dlopen weren't being found on FreeBSD
B CppMunge.py wasn't being invoked on Unix systems
B shm_open is only in librt on Linux
B Linux-only signal type constants in FXProcess are now compiled on Linux!
B Fatal and non retryable errors were disabling the wrong button in
FXExceptionDialog
B Fixed faulty referencing of a string in FXExceptionDialog (caused SIGBUS
on BSD)
B Fixed FXMutex::lock() causing abort when called from a cleanup/signal
handler on FreeBSD
B Fixed amazingly stupid mistake by me in FXThread::checkForTerminate()
where I was testing for thread cancel while holding the thread's mutex
B FXThread was not specifying PTHREAD_SCOPE_SYSTEM, thus it was creating
process-side only threads (these don't work and aren't consistent with Win32)
B Fixed bug on POSIX when opening FXMemMap with IO_ReadWrite actually
opened write only
B Fixed DLL unload failing to remove translations on POSIX (now TestDLL passes)
B FXACL::check() didn't account for root being able to do anything
B Fixed GCC compile errors in the patches to Boost for the python bindings
B FXHandedInterface::onCmdAccept wasn't invoking subclass message handler
B FXSSLPKey::operator>> was segfaulting
B Using symmetric keys bigger than 256 bits was corrupting the stack and
secure heap
B QPtrList::at, getFirst, getLast etc. were not returning zero if the list
was empty. Similarly for QPtrVector.

+ Added Generic::TL::findC
+ Added comparison operators to FXSSLPKey and FXSSLKey
+ Added FXSSLPKey::publicKeyAsString()
+ Added Secure::TigerHashValue::asString()
+ Added FXSSLPKey::publicKeyAsHash()
* Finally reimplemented FXFile in native Win32 calls
+ Added FXSettingsHelpers
+ Added FXACLEntity::isLoginPassword() with some difficulty
* FXTextField now zeros its contents on destruct if it holds a password
* Rewrote FXFSMonitor to use SGI's FAM on POSIX
+ Added FXACLEntity::owner()
+ Added inheritability support to FXACL
* FXACL can now set object ownership to entities other than the current user
* Reimplemented FXProcess::noOfProcessors(), mappedFiles(), hostOS(),
hostOSDescription() and hostOSMemoryLoad() to use native FreeBSD functionality
* Main SConstruct now builds tnfox by default
* FXThread now forces a thread stack of 512Kb
* FXThread now can schedule a thread either by the kernel or in-process and
adjusted FXFSMonitor and FXIPCChannel to be inprocess scheduled
* Updated python bindings to use Boost v1.31 and latest indexing suite v2 CVS
* Reembedded Spanish translations
* Added support for -fvisibility patched GCC which dramatically cuts down the
dynamic symbol table of the shared libraries, giving much improved link times
* Had FXFileInfo trap exceptions from FXACL, converting them into an empty ACL
instead
* It turns out that GCC will not transmit typed exceptions across a shared
object boundary unless they have a virtual function table
* Delayed signalling of wait condition for thread ended until after self
destruction
* Added QPtrList::takeFirst() and takeLast()

v0.7 9th March 2004:
-=-=-=-=-=-=-=-=-=-=
B FXString::number(double, fmt, prec) was defined but not implemented
B QDictBase::statistics was using spread as a variable and function
B QDictBase copy constructor forgot to copy hash table size
B FXThread cleanup and destructor were occasionally clashing (without locking)
B FXPrimaryButton wasn't actually generating SEL_APPLY messages :(
B Generic::Traits was choking on void *
B Generic::Functor was not trapping std::bad_alloc and was missing operator=()
B FXDataTargetI const operator * wouldn't compile
B Fixed window reparenting causing segfault (FOX bug)
B Fixed window reparenting not reparenting popup windows (FOX bug)
B Fixed timers, async messages and idle processing not being issued when
Windows is inside a modal loop (FOX bug)
B Clarified operator=() in FXDataTargetI::Returner
B Two small bits of malcode in QDictBaseIterator fixed
B Fixed up FXAtomicInt & FXMutex's use of inline so GCC v3.4 would be happy
B Malcode in QPtrVector, QPtrList, QValueList and QMemArray fixed
B QDictBase::resize() wasn't changing size causing dictionary corruption
B FXPipe::size() was failing with error on Win32 if the pipe hadn't been
connected to yet
B FXHostAddress::toString() returned the eight bit parts in reverse order :)
B Fixed bug where if a cancellable thing was done during cleanup the thread lock
never got freed
B Fixed bug where posting async messages on POSIX deadlocked
B FXPipe::writeBlock() wasn't a thread cancellation point on Windows
B FXLRUCacheIterator was using non-public types from FXLRUCache

* Tuned changes when BUILDING_TCLIENT is enabled
* Made Generic::isConvertible disable MSVC warning about possible bits lost
+ Added Generic::TL::dynamicAt
* Removed FXPipe isMsgType, added isDeepPipe and adjusted everything else
to suit
* Enhanced FXTransString so it can return a translation for any language
+ Added Generic::TL::dynamicAt
* Modified scons build system to let Tn build it remotely
* Replaced BUILDING_TCLIENT with BUILDING_TCOMMON. Not sure why I chose
TClient in the first place, it wasn't right as the Tornado client DLL
doesn't really need to exist anymore
+ Added fxstrfval()
+ Added operator << and >> for bool in FXStream
* Added SEL_RESET to FOX. Only FXPrimaryButton supports it yet
+ Added FXApplyResetList
* Added load() and store() to FXPrimaryButton
+ Added FXHandedDialog
* Moved FXExceptionDialog over to use FXHandedDialog
* Added upcall functors to FXDataTarget
+ Added FXFunctorTarget
* Moved everything accepting a Generic::Functor to accept by value instead
of reference
* Rewrote FXHandedDialog as FXHandedInterface and added FXHandedPopup
* Added copy() to FXTransString
+ Added screen scaling and screen constraining
* Fixed both FXExceptionDialog and FXHandedInterface which were using fixed
padding
* Hatched timers, chores and async msgs out into separate functions in
FXEventLoop
+ Added FXEventLoop::requestAsyncRecalc()
* Added Generic::TL::applyInstance and replicate
* Lots of daft little changes to gain GCC v3.4 compatibility. Probably worth
it in the long run.
+ New compiler=<tool> switch for scons file
* Removed the const PtrRef() function for Generic::ptr as technically it was
unsafe.
+ Added Generic::addConstRef
* Improved VersionedSharedLibrary in build tools
+ Added Intel C++ compiler for Linux support
+ Added QQuickList
* Added FXIPCMessagingBase
+ Added Generic::MapBools
+ Added strongestAnonCipher() and fastestAnonCipher() to FXSSLDevice
+ Added Generic::TL::filter and renamed applyInstance to apply
* Slightly adjusted the API of FXIPCChannel
+ Added waitForData() to FXIODevice
+ Added Secure::PRandomness
+ Added operators << and >> to Secure::TigerHashValue
* At long, long last FXBlkSocket::waitForConnection() is a proper thread
cancellation point on Windows. Also added a timeout.
* Merged fox v1.1.48
* Added support for getting memory page locking to work via a setuid process
* Updated python bindings for latest API
* Reembedded Spanish translations

v0.6 7th February 2004:
-=-=-=-=-=-=-=-=-=-=-=-
B QIntDictIterator didn't compile. Surprised I hadn't seen this before.
B FXStream::byteOrder() and setByteOrder() inverted on big endian machines
B Two TnFOX based processes were not playing nice on Windows
B FXStream was always swapping bytes by default irrespective of host endian
B FXResourceException was not passing through the flags so throwing a
FXConnectionLostException was very slow cos of the debug report production
B FXProcess::mappedFiles() now always returns module path - now FXTrans
works correctly on Linux
B On Windows FXBlkSocket readBlock() and writeBlock() were saying they had
completed when in fact they had not and the data transfer was happening
some time afterwards
B FXSSLDevice was faulting if the executable name was more than 16 or so chars
B SConstruct files for python and tests weren't setting FOX_BIGENDIAN
B FXSSLDevice didn't prevent multiple threads negotiating the connection
concurrently
B FXSSLDevice treated zero bytes read and written as an error
B FXSSLDevice::close() called twice caused exception
B FXGZipDevice compressed too much data if closed then opened
B FXWaitCondition now correctly uses ETIMEDOUT (thanks to Matthew Robertson)
B Generic::ptr when doing if(ptr==type *) was selecting wrong operator overload
B Fixed small bug in QSortedList::findIter which prevented compiling
B Fixed long-standing bug that operator new threw an exception on failure to
allocate. Now returns zero.
B QPtrListIterator was not always returning zero when invalid
B FXProcess static init was sending an iterator behind the first element
B FXMemMap wasn't resetting next mapping iterator when the mappings were none
B FXLocalPipe::readBlock() was returning before maxlen read
B FXLocalPipe::readBlock() was reading too much data (causing memory corruption)
B Discovered that you cannot define operator new in a namespace under ISO C++
(despite that both MSVC and GCC let you). Reworked code to use per-object
operator new & delete's.
B Major bug since maybe v0.4 of TnFOX - XIM hangs if X Threads are enabled.
Now disables XIM.
B Python StopIteration exception now specially handled by C++ side
B FXPythonInterp destructor was causing fault if called via FXThread::cleanup()
B Fixed last thread unlock when unwinding from FXProcess destroy faulting
B FXACL on POSIX was only writing other bits for everyone permissions, not also
user + group
B On POSIX error handling was using function return code (always -1) instead of
POSIX errno global
B QSortedList was inserting items into the wrong place if it were at the end of
the list
B FXException destructor segfaulted if the nested exception handling state had
been deleted (GCC keeps exceptions round longer than MSVC)

* Added true colour and alpha support to FXCursor. Submitted changes to FOX so
it'll likely appear there soonish. Added TestCursor.
+ Ported FXIPCMsg, FXIPCMsgRegistry, FXIPCMsg_ErrorOccurred, FXIPCChannel from
Tornado.
+ Added FXIPCMsgChunkCodeAlloc, FXIPCMsgChunk, FXIPCMsgRegister,
FXIPCMsg_Disconnect and FXIPCMsg_Unhandled
* Added TestIPC
* Improved error message handling in FXSecure entropy gatherer
* Added myId() to FXThread
* Added hostOSProcessorLoad(), hostOSMemoryLoad() and hostOSDiscIOLoad() to
FXProcess
* FXThread::id() now uses pthread_self() on POSIX
* FXExceptionCodes is now an enum (for python bindings)
* Added system exception FXNotSupportedException
* Added FXIODevice::readBlockFrom() and writeBlockTo(). Implemented for
all devices where it was appropriate.
+ Added fxdump8() and fxdump32()
* Added Winsock error code to error string decoding to FXBlkSocket on Win32
* Merged FOX v1.1.43
* Added on-demand thread creation to FXThreadPool
* Completely reworked FXRefedObject.h again :), losing FX::Pol::lockable
and kin and now using Generic::ptr as a base
* Some changes made to Generic::ptr
* Added missing operator[] to all QTL dictionaries
+ Added FXLRUCache, FXLRUCacheIterator and QCache, QIntCache specialisations.
* Added IO_ShredTruncate to FXIODevice and shredData(). Added same to all file
i/o classes
* Added yield() to FXThread
+ Added FXMemoryPool
* Added fatal exit upcall to FXProcess, plus FXProcess_MemLock
* On release builds of TnFOX on MSVC, now uses FXMemoryPool for all memory ops
* FX::Secure heap now uses a FXMemoryPool locked in memory and destroyed on
process exit no matter what
* Added FXMutex::setMutexDebugYield()
+ Added FXAPIR to restrict exports of code unsuitable for Tn components and
made a number of obsolete classes (such anything handling a directory) compile
nothing when BUILDING_TCLIENT is defined. Expect this list to increase substantially
with time.
* Hived FXEventLoop out of FXApp. Made various methods virtual. Added
getEventLoop() to FXId. Adjusted FXPopup and FXWindow to use event loops.
+ Added FXLockable, FXLockHold and made FXApp and FXEventLoop inherit from them.
* Added FXEventLoop::postAsyncMessage()
+ Added FXEventLoop_Static
* Split getNextEvent() into that and getNextEventI(). Merged Win32 and X11
getNextEvent(). Split peekEvent() into that and peekEventI(). Merged Win32 and
X11 peekEvent().
* Made FXSSLDevice no longer import the Secure namespace (GCC just didn't like
it :( ). This should hopefully fix the GCC segfault problems.
* Removed debug printing from FXSSLDevice in release mode (whoops!)
* Made FXSSLDevice use the secure heap for temporary buffers
* Made FXSettings threadsafe by locking through FXApp
* Added FXEXCEPTION_FOXCALLING1 and FXEXCEPTION_FOXCALLING2
* Added Xft2 detection & support to scons config
* Added FXDataTargetI
* Added operator[] to QPtrVector
+ Added FXPrimaryButton and made all appropriate FOX classes use it
* Added userHandedness() to FXProcess and userHandednessLayout() to FXWindow
* Renamed FXExceptionBox to FXExceptionDialog
* Improved FXExceptionDialog plus added handedness
* Added SEL_APPLY to FOX. Only FXPrimaryButton supports it yet
* Added user handedness and memory full overrides to the command line options
* Reembedded spanish translations

v0.5 24th December 2003:
-=-=-=-=-=-=-=-=-=-=-=-=
B Fixed invalid memory access when default FXProcess constructor used
B Fixed tiny bug where Generic::typeInfo wasn't demangling symbols on GCC
B QDictBaseIterator was just totally broken :(
B :: in class specifiers in translation files were being misinterpreted as line ends
B Translations registered by dynamically loaded DLL's weren't being unregistered
B iterators in some calls in QPtrList were being treated as random-access
B QDictBase::spread() didn't compile when instantiated
B Fixed two year old issue regarding setting the correct thread return code
B Fixed 40 byte per FXThread memory leak
B Fixed thread deletion causing disabling of nested exception framework (erk!)
B Fixed thread deletion destroying its exception TIB only to recreate it again.
B Finally, after only four releases, fixed the memory leakage in FXTrans
B Fixed bug in FXWaitCondition::wakeOne() whereby calling it when nothing was
waiting caused the condition never to signal.
B QValueList::remove(it) wouldn't compile
B %% in FXString arg() inserts was causing infinite loop
B FXString::nullStr() wasn't defined
B QDict wasn't handling case insensitive keys
B Fixed memory leak where FXTransStringArg didn't have virtual destructor
B QValueList::remove(iterator) was needlessly searching the list :(
B Fixed memory leaks caused by FXProcess exiting early
B Reworked Generic::convertible as GCC was getting confused during nested
Traits determinants.
B Primary (faked) FXThread wasn't setting FXThread::current() to zero after
destruction.
B If FXThread::run() terminated via return, cleanup wasn't being called on
Win32 only (with Win32 API used directly)
B Was failing to lock FXThreadPrivate during initial stages of thread execution
B Calling FXThread::enableTermination() without a prior disableTermination()
caused misoperation

+ Added the asList type to Generic::FnInfo
+ Added instantiateH and instance to Generic::TL
+ Added Generic::Functor, Generic::BoundFunctor
+ Added Generic::BindFunctor, Generic::BindFunc and Generic::BindObj
* Renamed FXDoUndo to Generic::DoUndo and added support for bound functors
* Replaced FXProcess::dllResolve() with functor based code
+ Added Generic::FnFromList
+ Added test of DLL facilities + dynamic loading/unloading of translations
* Replaced FXThread::addCleanupCall() with bound functor based code
+ Added FXThread creation upcall
+ Added QPtrVector and QPtrVectorIterator
* Replaced various routines which used QPtrList with QPtrVector
* Improved FXTrans' matching of translations to their provider DLL (removed
hard 10Mb DLL size limit).
* Added languages provided plus translation table stats to -sysinfo output
+ Added a policy-based smart pointer, FX::Generic::ptr
+ Added FXThreadPool as well as FXProcess::threadPool()
+ Added FXACLEntity, FXACL, FXACLIterator
+ Added FXFileInfo, FXFileInfoIterator
+ Added a policy-driven generic QValueList sorting class
+ Added FXDir
+ Added FXFSMonitor
+ Added QStringList
* CppMunge now prints file and line number if it crashes
* Added FXACL support to FXFile, FXPipe, FXMemMap, FXBlkSocket
* FXERRGOS() and FXERRGWIN() have been reworked to call a static method
rather than generate the exception inline (substantial bloat reduction).
Furthermore throwing of custom exception types are now more consistent
between platforms.
* Added the custom exception type "NoPermission"
* Modified FXFile to create its own file handle as you must specify then
if you might want to alter the DACL later
+ Added fxerrhpy() and FXERRHPY() plus custom python exception and
translation
* Rewrote FXPythonInterp
+ Added FXCodeToPythonCode plus defaults for FOX sort functions
+ Added FXPython::evaluate() plus py() and pyeval()
* Folded the BPL into TnFOX.pyd
* Changed FXThread::checkForTerminate() to return bool
* Updated Spanish language translations
+ Added policies for new TnFOX classes
* Updated SConstruct files to work with scons v0.94
* Embedded spanish translation file
* Changed FXProcess::dllLoad() to optionally append .so and prepend lib

v0.4 10th November 2003
-=-=-=-=-=-=-=-=-=-=-=-
B Fixed FXHostAddress where I'd used the wrong IPv4 to IPv6 mapping!
B Major bug this - FXRollback when taken was exiting process due to pure virtual
call. Whoops!
B FXBlkSocket::accept() now supports thread termination
B Fixed small for() scoping problem in FXException
B FXThread::sleep() and FXThread::msleep() now support thread termination
B Fixed two places in FXPipe where it wasn't handling pipe broken correctly on Win32
B Fixed UpdateMunged.py not piping stderr to its stdout
B Fixed CppMunge.py fetching source file name from variable which no longer exists
B Fixed FXVisual missing disabled copy constructor
B Stopped memdbg warnings on GCC
B Implemented FXStream::getSpace and setSpace (forgot to implement them)
B FXString comparison operators were returning FXbool, not bool
B FXProcess::dllLoad didn't find DLLs in the same directory as the DLL being loaded

+ Moved entire project to work out of subversion mainly cos CVS sucks
+ Added FXRBConstruct(), then to all relevent parts of the TnFOX extensions
+ Added FXTrans::provided() which I unfortunately defined but never implemented :(
+ Added FXDoUndo plus several other generic programming tools (into FXGenericTools.h
and FXPolicies.h)
* Made FXRefedObject use policies (still source compatible)
+ Added the FX::Secure namespace
+ Added FX::Secure::TigerHash
+ Added FX::Secure::Randomness
+ Added FXShrdMemMutex
+ Added FXIODeviceS and moved the relevent devices over to use it
* Added global new & delete operator overrides to FX namespace which never throw
+ Added FXSSLDevice, FXSSLKey & FXSSLPKey
+ Made all FOX header files standalone
+ Implemented return policies for all TnFOX methods needing them in bindings
* Moved build process over to scons
+ Added a large range of generic tools (too many to list here)
+ Added a mostly-complete set of python bindings for TnFOX
+ Produced patch file for same bindings
* Replaced QDictBase with completely new optimised implementation (and thus
QDict and QIntDict as well).
* Merged fox v1.1.38 + added relevent new policies
+ Added QPtrDict
* Altered license declaration to state specific LGPL version across all files
+ Added Python Test
+ Added Generic Tools Test
+ Added DLL binding to speed loading on Win2k SP3
+ Added custom libtool compatible output scons builder for GNU
* Embedded fully updated spanish translations
* Merged fox v1.1.41

v0.3 27th July 2003:
-=-=-=-=-=-=-=-=-=-=
B Fixed bug where FXIODevice::removeCRLF() was introducing ASCII 10's if the
blocks fed to it split mid-CRLF.
B Fixed bug where FXIODevice::applyCRLF() was not adding ASCII 10 in a CR,LF
sequence if output buffer was one too short.
B Fixed bug where CppMunge.py was mangling error macro extraction on complex
FXERRH()'s
B Fixed bug where FXThread internal thread cancellation event wasn't being
created for primary thread
B Fixed bug in FXZeroedWait where wait condition and count were becoming desynced
B Fixed bug where operators += and -= in FXAtomicInt weren't returning the right value
B Fixed bug where FXThread::wait() with an infinite value was deadlocking on POSIX
B Fixed resource leak when a FXThread was destroyed on POSIX
B Fixed FXThread::enableTermination terminating the thread immediately :)

+ Made FXERRH_ISINFORMATIONAL exceptions or those without source file id not
include source file nor stack backtrace info in the error report
+ Altered FXERRHWIN() in WindowsGubbins.h to specially trap file and path not
found errors and generate a FXNotFoundException instead.
+ Added code to FXProcess to generate a secure default DACL for inter-process
used objects.
+ Added FXPipe, plus anonymous pipe support
+ Added FXHostAddress
+ Added FXBlkSocket
+ Added FXLocalPipe
+ Added QSortedList
+ Added pageSize() to FXProcess
+ Added FXMemMap
+ Added FXNetwork, both new-style DNS lookups and old cos Windows' new styles
are broken!
* Stopped CppMunge.py extracting FXEXCEPTION_* macros
* Added trapping of memory allocation failures to QTL implementations
+ Added an adler32 checksum calculation routine, thanks Jeroen!
* Added lots of new tests to TestDeviceIO
* Merged fox v1.1.32
+ Generated a full set of project files for MSVC7.1 (.NET 2003)
* Got rid of a lot of the GCC warnings about the QTL
+ Finally got the dllResolve() templated function pointer syntax right in FXProcess
* Harmonised Windows and Linux thread cancellation behaviour
+ Ported FXRefedObject classes from Tornado
* Embedded latest spanish translations of string literals

v0.2 12th July 2003:
-=-=-=-=-=-=-=-=-=-=
B Fixed bug where copy to clipboard in error reporting box wasn't working on X11
B Fixed bug where FXProcess' destructor was not destructing static inits
B Fixed bug where FXThread was not calling primary thread registered cleanup calls

+ Added string literal extraction to CppMunge.py & generation of translation file
* Merged fox v1.1.29
+ QDict, QDictIterator, QIntDict, QIntDictIterator added
+ Altered xincs.h to use 64 bit file addressing on both Unix & Windows + as many API's
as I could find now use FXfval to indicate file size and position
+ Added FXIODevice
+ Added FXFile
* Revamped FXStream:
  + Now uses an FXIODevice for i/o
  + Added much more error detection
  + Carefully changed to maintain backwards API compatibility
* Altered FXFileStream to thunk to FXStream + FXFile combination
+ Added QMemArray<> and QByteArray
+ Added FXBuffer
* Altered FXMemoryStream to thunk to FXStream + FXBuffer combination
* Altered fxgifio.cpp, fxjpegio.cpp, fxpngio.cpp, fxrgbio.cpp, fxtifio.cpp &
fxxpmio.cpp to be compatible with new 64-bit i/o structure
+ Added utext() and wtext() to FXString for eventual unicode transition.
+ Added FXGZipDevice, removed FXGzStream
+ Added FXStream overloads for all the QTL thunks, FXException, FXBuffer.
* Merged fox v1.1.30
+ Added FXMemDbg, plus added it to all my source files.
+ Added generic FXStream overload for all FXIODevice's
+ Added stdio() method to FXFile (really needs FXPipe)
+ Added transaction rollback support (FXRollback*)
+ Implemented FXTrans plus added FXTransString
+ Moved to new TnFOX MSVC project, changed file extensions to .cxx, installed
custom build configuration to have CppMunge.py process all C++ source before compilation.
+ Added dynamically loaded library support to FXProcess and FXTrans
+ Generated MSVC7.1 project files for all parts
+ Added -sysinfo on the command-line
+ Added FXProcess::mappedFiles() which now means FXTrans knows where to find its
translation files
+ Added automatic GZip decompression and compression support to FXTrans
+ Added delayed DLL loading on Win32 for mostly unused DLL's
+ Added StripSymbols to generate .dbg files for MSVC release builds
+ Merged fox v1.1.31

v0.1 19th June 2003:
-=-=-=-=-=-=-=-=-=-=
First version released to public