Windows2000/private/windbg64/debugger/ee/changes.txt

11-Jun-1996 a-MatH
		Allowed the ee to work for multiple target machine types.  Removed any
		#ifdef TARGMACPPC &c and changed to if (TargetMachine == mtpmppc).
		
15-Jan-1996 BryanT
        *.c *.h sources
	        Use .hxx versions of ee/sapi/cvtypes headers.  Don't need ldouble.c.

10-Jan-1996 BryanT
        *.c *.h
	        Make it compile/link as C++ code.  This uncovered a number of problems
		with functions w/o return values, incorrect enum usage, etc.

09-Jan-1996 BryanT
        *.c *.h
	        Delete Win16/DOS support.  Detab.  Format to OSDEBUG standard.
		Replace _fmemxxx with memxxx.  Replace _ftcsxxx with _tcsxxx.
		Remove PASCAL, FAR, NEAR, LOADDS, LOCAL.
		Remove C_ONLY and ADDR_xxx usage (the EE is always C++ ADDR_MIXED).

08-Jan-1996 BryanT
        *.asm *.inc *.lib
	empty.c mat.c libmain.c
		Delete Win16/DOS files.

	sources makefile eecxx.def eecxx.rc
	        Add NT style build process.

07-Aug-1995 SanjayS
	debfmt.c dllmain.c
		Support automatic display of unicode strings/arrays.
03-Jul-1995 SanjayS
	debtyper.c
		Don't allow arithmetic on void pointers.
28-Jun-1995 Sanjays
	debapi.c debsup.c debfmt.c debsym.c
		Allow settings bps on bound virtual functions. We figure out the derived most class
		and construct the name of the func preceded by is class scope i.e. CFoo::vfunc and
		then try to set a bp on that.
27-Apr-1995 Sanjays
	debeval.c
		Fix virtual func ptr evals. We read the vtable ptr as a sym (4 byte value)
		and need to take this value and work out the actual unfixed up address.
30-Mar-1995 Sanjays
	debsup.c debfmt.c
		Support multiple level auto-expand rules. Support -> in
		autoexpand strings.
30-Mar-1995 v-LeoL
	debsup.c
		Special case __$ReturnUdt in SetFcniParm for pretty printing
		function signature

21-Mar-1995 Sanjays
	debsup.c
		GetParmTM did not intialize pExState like is should.	

22-Feb-1995 v-LeoL
	debbind.c debeval.c shfunc.h
		MIPS Function evaluation
	debexpr.h debsym.c
		Cleanup
	eent.mak
		Use msvcrtd for DEBUG build

10-Jan-1995 v-LeoL
	debeval.c debbind.c
		Support for $ReturnValue on MIPS
        ldouble.h
		Use pragma pack(push/pop)

06-Jan-1995 v-LeoL
	debsym.c
		Tune fValidInProlog to toss out only SP/S8 regrel symbols
        debsrch.c
		Olympus 3763 -- Fix to avoid Float10 use on MIPS

14-Dec-1994 v-LeoL
	debsym.c
		fValidInProlog assumes that offsets are FP relative --
		MIPS are SP relative so return FALSE for all symbols.
		[An optimization would be to pass in frame size and then
		test offset >= framesize+16]

13-Dec-1994 v-LeoL
	deblex.c
		L"a" was getting lexed as IDENT STRING. moved Ident test
	        after string test and added L'a' support.

16-Nov-1994 SanjayS
	debsym.c
		Cull search at first hit for Auto expand in Multiple inheritance scenarios.
	debwalk.c
		Incorrect handling of MBCS filenames.
	debutil.c
		Fix Olympus:2864. Change check for type validity.	
07-Nov-1994 SanjayS
	debsup.c
		Speed up GetMemberIA for the common case.
03-Nov-1994 SanjayS
	debsup.c
	debapi.c
	debexpr.h
		Added EEGetBCIA to support the this window in the IDE.
11-Oct-1994 RicoM
	debsup.c
		pCxt was not initialized while getting the ith child of
		a class if the class reference variable.  Olympus #1723

23-Sept-1994 RicoM
	debutil.c
	debfmt.c
		Revisiting Dolphin Bug #14596.  Now using the context from
		the expression to find UDT symbols.  We had been trying to
		make a context from the HMOD but this turns out to be unsafe
		see debutil.c

21-Sept-1994 RicoM
	debbind.c
	debexpr.h
	debsym.c
		Allow binding to labels inside of functions even if the
		frame for that function has not been instantiated yet
		(this regressed when we tryed to not let you bind to
		arguments of functions that were not yet instantiated)

02-Sept-1994 GeorgioP
	debbind.c
	debexpr.h
	debsym.c
		Support display of function arguments while being in the prolog
		of the function.

31-Aug-1994 RicoM
	debbind.c
		Minor fix to auto-expand to avoid re-freeing memory

31-Aug-1994 GeorgioP
	debfmt.c
		Low budget unicode display via the ",mu" and ",su" format specifiers

25-Aug-1994 GeorgioP
	debapi.c
	debbind.c
	debeval.c
	debexpr.h
	debfmt.c
	debparse.c
	debsup.c
	debsym.c
	debsym.h
	debutil.c
	dllmain.c
		Support auto-expansion of class members. Each TM that is a class or
		a pointer to a class has now a list of auto-expanded child TMs that
		are created and displayed based on expansion rules that reside
		in the "AutoExpand" section of "msvc.ini". (E.g., a rule like
		CPoint = x=<x>,y=<y> allows displaying points like
		point = {x=1,y=2} instead of point = {...})

27-Jul-1994 GeorgioP
	debbind.c
	debeval.c
	debexpr.h
	deblex.c
	debparse.c
	debsrch.c
	debsym.c
	debutil.c
	debwalk.c
	expr2.z
		Support the "$ReturnValue" keyword that binds to the return value
		of the current function (assuming the current function is ready
		to return). This feature enables the kernel (or even the user) to
		access the return value of the current function. The kernel can
		also get the return value of a function that has just returned, by
		using an appropriate context.
		Implementation (x86 only): A new operator, OP_retval, has been
		added in order to handle "$ReturnValue" as a special identifier.
		If the return value resides in a register, the OP_retval node
		is marked as a register value. The reg field that used to be a
		union member in the evaluation node is now a separate struct, so
		that register-storage information do not overwrite type
		information. Register-handling code has been extended to handle
		registers having non-trivial types.
	

11-Jul-1994 RicoM
	debutil.c
		Bug #14596.  Fixed SetDPtr so that it doesn't fault if it can't find
		the type that it is looking for in the module that it's supposed
		to be in.  This now "masks" a more serious problem that
		forward reference type need to be looked up using the context that
		is in the evaluation node (when HMOD != NULL) instead of the global
		context (which may be totally inappropriate for that node).
		Also, a fake node and context should be created when formatting pointers
		to functions instead of just faking the HMOD.

16-Jun-1994 Georgiop
	debsrch.c
		Apply this adjustment at the end of LoadAddress [Dolphin #13652]
		(Code under #if THISADJUST, until the compiler bug gets fixed)
	debexpr.h
	debsup.c
		Fixed counting of overloaded methods in GetClassiChild.

13-Jun-1994 Georgiop
	debbind.c
	debexpr.h
	debsup.c
		Addition to the previous change: since the cost of EEcChildrenTM
		is now higher (due to potential presence of static members) and
		since the kernel calls EEcChildrenTM quite often (via
		LTIsExpandableTM), we now save the child count in the TM.
		Also BindStMember now saves and restores pCxt

12-Jun-1994 Georgiop
	debbind.c
	debexpr.h
	debsup.c
		Do not display non-present static data members when expanding
		a class TM. EEcChildrenTM and EEGetChildTM ignore static data
		members that are not present.
		In addition search state for GetClassiChild is saved in the TM
		so that search for child (i+1) can continue from the point where
		search for child i ended.

31-May-1994 v-LeoL
	debsup.c
		Avoid using pCxt in ResolveAddr. When called from EEInfoFromTM
		this is undefined. Use pExState->cxt instead.

19-May-1994 Geogiop
	debexpr.h
	debsym.c
	debwalk.c
		The context list returned by EEGetCXTLFromTM contained a starting
		context from which the corresponding symbol was visible, not the
		exact context where the symbol was found. The kernel needs to know
		the exact context for each symbol in order to set breakpoints
		properly. I extended eval_t to accomodate full context information,
		so that SearchSym can store the context where a symbol was found.

14-May-1994 Georgiop
	debbind.c
	debeval.c
	debfmt.c
	debsrch.c
	debtyper.c
	debutil.c
		Provided support for T_(U)INT1 types (FORTRAN uses them)
		[Dolphin #12021]

12-May-1994 Georgiop
	debtyper.c
		CastNode was not setting ADDR_IS_OFF32 properly, and PushUserValue
		ended up pushing 2 bytes instead of 4 for function arguments that
		were cast to ptrs. [Dolphin #12157]

06-May-1994 Georgiop
	debexpr.h
	debparse.c
	debapi.c
	debsup.c
		Allow transformation of template names to a 'canonical' form only
		when EEParse is called. In other cases (e.g., when ParseBind is
		called) the expression should have been generated by the EE and
		should be in canonical form already. This optimization makes the
		EE more tolerant to debug info idiosyncracies as we avoid
		transforming template names that have been generated by the
		compiler (even if these are not really 'canonical' as far as the
		EE is concerned).               [Dolphin #11696]

04-May-1994 Georgiop
	debbind.c
	debeval.c
	debexpr.h
	debfmt.c
	deblex.c
	debparse.c
	debsrch.c
	debsym.c
	debtyper.c
	debutil.c
		Enhanced support for __int64: common binary and unary opearations
		can now accept int64 operands, and int64 string conversions have
		been enabled. Floating point <-> int64 conversions are not supported
		yet in order to avoid build problems due to potential unsupported
		operations by the compiler. Code involving pointer segments and
		offsets does not treat int64 in a special manner.

25-Apr-1994 v-LeoL
	shfunc.h debsup.c
		Support for locals inside nested procedures. C++EH Catch blocks
		and SEH finally blocks and SEH filters are implemented via
		nested procedures on MIPS
		
22-Apr-1994 Georgiop
	debsym.c
		AddVBType was not updating "*ppList" and "pList->MaxIndex"
		following a reallocation of "*phList". This could eventually
		cause a heap corruption and a GPF. [Dolphin #10293]

21-Apr-1994 Georgiop
	debbind.c
		Disabled overloaded operator functionality for "&". The EE uses the
		non-overloaded "&" operator internally for constructing child
		expressions. When rebinding nested child expressions the EE would
		waste a lot of time looking for potential overloads of "&". As a
		result the watch window would become extremely slugish and the
		IDE would appear to be hung. [Dolphin #10293]

20-Apr-1994 Georgiop
	debbind.c
		BindSymbol prefers binding to a non-UDT symbol in case both a UDT and
		a non-UDT have the same name and appear in the same scope. (Most of
		the time, the user really wants the variable, not the type)
		[Dolphin #10794]

18-Apr-1994 v-LeoL
	debeval.c debbind.c
		Change error for Mips func eval from ERR_CALLSEQ to
		ERR_NOFUNCEVAL

15-Apr-1994 v-LeoL
	debeval.c
		Fix warning
	debtyper.c
		Another place to call ResolveAddr
		
15-Apr-1994 Georgiop
	debeval.c
	debexpr.h
	debfmt.c
	debsrch.c
	debtyper.c
	debutil.c
	fmtstr.h
		Provided minimal support for __int64, as the EE would hang or
		misbehave while handling this type. Most operations on __int64
		are not supported yet, but __int64 variables can be evaluated
		and displayed in hex.  [Dolphin 10611]

11-Apr-1994 Georgiop
	debfmt.c
	debutil.c
		[Dolphin 10456]

		FormatExpand now handles a SetNodeType failure that could set
		the underlying type of an array to T_NOTYPE and cause a divide
		by 0 exception.

		SetNodeType resolves fwd references to underlying array types
		while setting the node of an array type. This is done in order to
		avoid having to resolve the type at a later time, when the
		appropriate context for the symbol search may be unavailable.


07-Apr-1994 Georgiop
	debtree.c
		AddLeaf and AddNode were not checking for stack space in the
		syntax tree before adding a new node. This could cause a memory
		corruption and a GPF, since the the tree stack size is fixed
		(although the rest of the tree may grow). [Dolphin #10293]

05-Apr-1994 v-LeoL
	debbind.c debeval.c debtyper.c
		Fix uninitialized local (Dolphin 7755) and enable some MAC/PPC
		ifdefs for MIPS too.

04-Apr-1994 v-LeoL
	debsym.c
		If we find something in the publics don't save uninitialized
		CXT [fix is MIPS only for now]

01-Apr-1994 Georgiop
	debsup.c
		Changed GetNameFromHSYM to skip the numeric leaf before accessing
		the symbol name. The "value" field of CONSTSYM has fixed length
		at compile time, and is followed by the "name" field. In fact
		CONSTSYM is a numeric leaf that may have variable length, and
		we ended up treating the last bytes of the numeric leaf as part
		of the symbol name. [Dolphin #9323]
	
	debtyper.c
		CastNode now sets the ADDR_IS_FLAT flag. (EvalMem was emitting
		a segmented address after calling CastNode.) [Dolphin #9941]

	debfmt.c
		Do not allow garbage following a format specifier; allow only
		trailing white space. [Dolphin #9902, #9903]

	debexpr.h
	debbind.c
	debutil.c
	debeval.c
		Added routine for checking if pointer subtraction is allowed, as
		the existing check for type equality was too restrictive.
		[Dolphin #8914]

		
01-Apr-1994 MarkBro
	debfmt.c:
		Changed size of buffer in EvalMem (bigger) and now handle
		buffer (in) is smaller than local copy of buffer (DOLPHIN:9942)

31-Mar-1994 v-LeoL
	debutil.c
		Protect against unaligned numeric leafs
		Also avoid Float10 when updating REAL32 and REAL64
		
30-Mar-1994 v-LeoL
	debsup.c
		Handle REGREL in InfoFromTM (use ResolveAddr to get BASE+off
		rather than just off)
		Also fix ResolveAddr to handle S8 correctly when at Top of Stack
		OSDEBUG4 does not return correct adrBase when alloca frame is
		topmost.
	
30-Mar-1994 Georgiop
	debwalk.c:
		OP_uscope now contains an eval_t node. BindBScope uses the eval_t
		node of its left child (which can be a unary scope operator) to
		set the CLASS_GLOBALTYPE flag. Since OP_uscope contained no eval_t
		node this was resulting to a memory corruption. [Dolphin 9185]

	debtree.c
		GetExpr was not skipping over embedded hSym values properly and
		returned garbage, preventing editing of child TMs in the locals
		window. [Dolphin 9842]

25-Mar-1994 Georgiop
	debsym.c
	debeval.c
		MatchArgs and EvalFunction were not handling fwd references to enum
		types. [Dolphin 3659]

24-Mar-1994 Georgiop
	debfmt.c
		Fixed formatting of const/volatile type qualifiers. (A type like
		"const Foo const *" was being displayed as "const const Foo *")
		[dolphin 5043]

23-Mar-1994 v-LeoL
	debeval.c
		As the EE knows nothing about how to make a MIPSCALL function
		call use ERR_CALLSEQ

16-Mar-1994 Georgiop
	debexpr.h
	debsym.c
	debutil.c
		Worked around the following bug:
		While self debugging, bp at LTIsExpandableTM and watch
			"{,,een0cxxd.dll}(pexstate_t*){*}phtm"
		The resulting expression is not expandable, because
		while in EEIsExpandable, we call SetNodeType and try to
		resolve a forward reference by performing a symbol search
		in the default context instead of {,,een0cxxd.dll}

		Since we cannot always update pCxt before calling SetNodeType,
		resolve potential forward references in undelrying types
		when they are first introduced int the evaluation node.

04-Mar-1994 Georgiop
	debbind.c
	debeval.c
	debexpr.h
		Dolphin #5503: We GP-faulted while binding an expression like
		Foo::m_array[Foo::m_staticCount], because we needed to fake
		up a stack entry even if ST was not NULL.
		Now we always push a faked entry unless there is a preceding
		bnOp (".", "->", etc); in that case a class stack entry has
		already been pushed onto the stack.

	debsup.c
		Dolphin #8336: When dereferencing a "reference to pointer" we
		should dereference the underlying pointer (this is what
		EEIsExpandable implies when it returns EEPOINTER)

	debparse.c
		pExState->strIndexSav should not be set by XFormExpr, since at that
		time the actual parsing has not started and the value of strIndex is
		unknown.

01-Mar-1994 v-LeoL
	debsup.c
		Dolphin #7908. Intel avoid printing 'this' param by checking
		for negative offset. Centaur has 0 offset for this (or even positive)
		so special case it.

	debutil.c
		Dolphin #7945

	debeval.c
		Dolphin #7945. OSDSaveRegister currently broken on MIPS so
		avoid calling in EvalFunction if doing EvalProto
		(EvalProto returns anyway without restoring)
		
24-Feb-1994 Georgiop
	debexpr.h
		CLEAR_EVAL_FLAGS was only clearing out part of the flags structure.
		(the "bits" struct in vbits_t had grown beyond 32bits in size,
		and the clr field was too small for clearing all bits.)

	debbind.c
		Restricted the symbol search to class names only, when binding the
		left operand of "::". The EE would erroneously bind the class name
		at the left of "::" to the corresponding constructor and rewrite
		the expression by inserting the "this" pointer. (This in turn
		prevented binding static members like Foo::m_static when the
		"this" pointer was not in scope).


23-Feb-1994 Georgiop
	debdef.h
	deberr.c
	debexpr.h
	debfmt.c
	eent.mak
	eemc.mak
	errors.h
	resource.h (new file)
	debmsb.rc (new file)
	dllmain.c (new file)
		Moved error messages to string resources so that they can be localized.

21-Feb-1994 RicoM
	debsym.c
		We no longer check the type index in the globals for and exact
		match with the type index in the class for static data members.
		This is because sometimes it doesn't match (e.g. unspecified
		length arrays)

20-Feb-1994 Georgiop
	debapi.c
	debbind.c
	deberr.c
	debfmt.c
	deblex.c
	debparse.c
	debsrch.c
	debsup.c
	debsym.c
	debutil.c
	debwalk.c
	debexpr.h
		Changes to use tchar.h - DBCS enabling

16-Feb-1994 RicoM
	debsym.c
		If nested type forward ref cannot be converted into a real type
		then pretend we didn't find the symbol we were looking for in
		the nested type and proceed with the search instead of returning
		an OMF error and abandoning the search

15-Feb-1994 DaveS
	errors.h
	debeval.c
		Fix function eval for mac68k.  We still don't eval funcs, but
		now we will eval BPs and prototypes.  Also, I've added a new
		error message for attempting function eval.

10-Feb-1994 Georgiop
	debbind.c
	debeval.c
		Fixed expansion of function calls that return references to derived
		classes. Such function calls are now treated as lvalues (as per the
		C++ spec), and it is possible to bind child expressions (such as
		base classes) that need the address of the referenced object.

02-Feb-1994 Georgiop
	debexpr.h
	deblex.c
	debeval.c
		Dolphin #5663: Work around for a compiler optimization bug: replaced
		instaces of "(float)DoubleFromFloat10" with new inlined function
		"FloatFromFloat10".

	debutil.c
		Dolphin #5783: RemoveIndir should clear the EVAL_IS_LABEL flag.

04-Feb-1994 Georgiop [by RicoM]
	debeval.c
		Dolphin 5530: Handle subtraction of void* pointers (fail gracefully).
		Misc: Fixed bug in Store32 -- ADDR packet uninitialized
	debsym.c
		Dolphin 5465:
		Fail symbol searches on member functions that are not
		instantiated in the current .exe/.dll (i.e. if a class
		imports member function you must be in the scope of the
		function to set a breakpoint -- the kernel does this search
		automatically)

29-Jan-1994 Georgiop
	debapi.c
	debeval.c
	debexpr.h
	debsup.c
	debutil.c
	expr2.z
		Implemented evaluation caching mechanism: child TMs can now
		be evaluated incrementally by reusing the evaluated result
		of a parent TM.

27-Jan-1994 MarkBro
	debfmt.c:
		Restored _MBCS enabled code from previous whoops

	debparse.c:
		Change _MBCS #ifdef's to #ifNdef's _SBCS

26-Jan-1994 BillJoy
	debsym.c
		pulled MHOmfUnLock (hCFlag) inside the #if !defined(TARGMAC68K)


23-Jan-1994 RicoM
	debutil.c
	debsym.c
		Fixed calls to GetUdtDefnTIndex() so that the type index
		corresponds to the module that the type is being looked
		up in.  Fixes Dolf#5544

23-Jan-1994 RicoM
	debapi.c
	shfunc.h
		Added stub implementation of EEInvalidateCache per GeorgioP

21-Jan-1994 MikeMo
	makefile
		Added -WX: no warnings allowed.

12-Jan-1994 Georgiop
	debsup.c
	debbind.c
		Special handling of context operators when generating synthesized
		expressions. Unified context handling scheme.
		Minor bug fixes and misc. cleanup for can vs cxx builds.
	debexpr.h
	debsrch.c
	debsup.c
	debsym.c
	debwalk.c
		Misc. cleanup for can vs cxx builds.
	
22-Dec-1993 MikeMo
	debsym.c
		Dolphin #4678: realloc was the wrong size.

06-Dec-1993 Georgiop
	debutil.c
		SetNodeType was not restoring PTR_REG_IREG and PTR_REG_HIBYTE
		of an enregistered pointer.

04-Dec-1993 Georgiop
	debeval.c
	debbind.c
		Modified EvalFunction in order to fixe dolphin 1170: When a
		simple variable is passed by ref we need to pass a pointer
		to the variable (i.e., treat it in the same way we currently
		treat passing classes by ref).
		Modified Push?Args functions in debbind.c in order to restore
		the PTR_UTYPE macros (when handling args that are passed by
		ref) that had somehow disappeared.

	debsym.c:
		Updated comment header for SearchSym in an attempt to document
		some of its side-effects. Didn't change any code.

	debparse.c
		Changed the condition on which a special path is taken when
		parsing consecutive closing ')' of nested function calls in
		order to fix dolphin #3404.

		Changed FParseExpr to treat 'by', 'wo', 'dw' as identifiers if
		they follow a '.' or a '->' (in that case they are probably
		struct memebers, not commands). (dolphin #977)

	debutil.c
	debfmt.c
		TypeDefSize and Format now use the f32bit flag ti decide on
		whether a based pointer is 16-bit or 32-bit wide (there is
		no distinction between these in codeview information).
		(dolphin #979)

30-Nov-1993 MarkBro
	debsym.c:
		Corrected initialization of localeval in SearchRoot for
		nested symbols.  Was asserting since the previous initialization
		of localeval of *pvF didn't have an associated module

05-Nov-1993 RicoM
	debsup.c
		Corrected off by one error allocating space for derived most
		class cast

01-Nov-1993 RicoM
	debsup.c
		EECChildren now returns 1 for # children of a pointer to a
		primitive type

26-Oct-1993 Stevesm
	debapi.c
		We had to initialize some of the global variables in
		EEFormatCXTFromPCXT for NB10 restart with breakpoints.  Resolving
		forward reference UDTs caused us to invoke SearchSym.  We needed to
		initialize pCxt, pExState and various Eval Stack variables
		
30-Oct-1993 Georgiop
	The EE was displaying bogus values for missing static class members.
	Now if a static data member is missing the corresponding expression
	binds properly and its evaluation or expansion is prevented. Affected
	files:
	debsym.c:
		SearchQualName sets the HSYM of the eval node to 0 in order to
		indicate a missing static data member. It also sets a global
		flag in the TM to prevent expansion.
	debeval.c:
		StructEval returns an error in case an evaluation of a missing
		static data member or function is attempted.
	errors.h:
		added ERR_STMEMBERNP ("static member not present") error
	debexpr.h:
		added fNotPresent flag in the state field of TMs
	debapi.c:
		EEIsExpandable: do not expand expressions that contain missing
		static data members.

29-Oct-1993 Georgiop
	debbind.c:
		BindDot allows binding of expressions like {main,,}foo.bar where
		foo is not yet in context and bar is a (non-virtual) member
		function. This allows setting certain break points when restarting
		a debugging session [dolphin #2433]. This fix addresses part of
		a more general issue, i.e., what happens if part of an expression
		cannot be bound, but the expression itself is bindable (i.e., does
		not need all the information of the part that cannot be bound). In
		general the EE does not allow binding of such expressions -- this
		fix overrides this behavior.

20-Oct-1993 Georgiop
	debexpr.h:
	debeval.c:
	debbind.c:
	debapi.c:
		Backed out the change made on 10-15-93 (fPreventEval flag). Now
		EEDereferenceTM verifies that the downcast node is bindable. This
		involves an extra call to EEGetChildTM, but results in better UI.
		BindCast does not allow illegal casts that pertain to the class
		hierarchy (e.g., casting a vitr. base ptr to a derived class ptr).
		[dolphin #2751]


15-Oct-1993 Georgiop
	debexpr.h:
	debeval.c:
	debbind.c:
		Introduced fPreventEval in the TM state in order to prevent
		evaluation of certain expressions. This is used together with
		the auto class cast feature as follows: Neither DereferenceTM
		or cChildrenTM check the validity of the cast required to build
		the derived class node (it would be too costly to do so). If
		BindCast discovers an illegal cast it lets the expression bind
		properly (so that the kernel can query the expression name) but
		prevents the evaluation of the node (which would produce meaningless
		results)


15-Oct-1993 Georgiop
	debeval.c:
		Minor changes in VFuncAddr2 and StructEval in order to support
		evaluation of virtual functino addresses. These should fix
		dolphin bug #2223

14-Oct-1993 Georgiop
	Dolphin bugs 299, 2312, 2223
	debsup.c:
		UpdateMem was not resetting properly the EVAL_IS_BPREL flag.
		(dolphin 299)
	debeval.c:
		StructEval now produces an ERR_NOTEVALUATABLE instead of
		ERR_INTERNAL if VFuncAddress2 fails. VFuncAddress2 may fail if the
		object does not have a valid vfptr -- dolphin 2223.
		(The problem of computing the right address of a virtual
		function still exists)
		EvalAddrOf() was not handling references properly. RemoveIndir()
		was not handling CV_PTR_NEAR32 types.

13-Oct-1993 Georgiop
	debfmt.c:
		FormatCXT has been changed to append an argument type list
		to function names in context strings, so that the kernel can
		disambiguate overloaded functions.
	debbind.c:
		ContextToken() now accepts tokens that contain a parenthesized
		substring. This allows parsing of contexts generated by
		FormatCXT. The parenthesized part is ignored, since BindContext
		cannot deal with overloaded functions
	debbind.c:
		Bug fix in BindContext: Typing "{foo,,}bar" in the watch window
		before starting a debugging session would cause an assertion
		in sapi, as BindContext was attempting to call SHGetNextMod
		passing a NULL hExe. This call has been removed.

12-Oct-1993 Georgiop
	Bug fix for Dolphin #2109: attempt to cast a register to a UDT
	during evaluation of an assignment failed due to the NULL hmod
	in the evaluation node of the register.
	debeval.c:
		Check if ST has a null hmod before calling CastNode() in
		Assign().In that case use the hmod found in STP

08-Oct-1993 Georgiop
	Further enhancements to auto class casting: nested class support,
	pointer adjustment when casting from "pointer to base" to
	"pointer to derived"
	debapi.c:
	debexpr.h:
	shfunc.h:
		New API EEEnableAutoClassCast
	debparse.c:
	debbind.c:
		Modified ParseTypeName and ParseType to allow contexts in
		type strings. Added CastBaseToDeriv().
	debsup.c:
		GetDerivClassName generates a context together with the
		type name. Minor changes in the way the magic node is
		formatted

05-Oct-1993 RicoM
	Minor bug fixes
	debsup.c
	debapi.c
		Skip LF_MODIFIER nodes during auto class cast so that
		auto class cast works even on const or volatile pointers
		Don't ever use BIND_fSupBase as it prevents expansion
	    of members found in base classes in
	
01-Oct-1993 Georgiop
	Implemented automatic detection of the type of a class
	object pointed by a ptr, in case the ptr points to an
	enclosed base class object which has a vtable.
	debexpr.h:
		Added hDClassName in pexstate_t to hold the
		derived class name.
	debsup.c
		Added GetDerivClassName, modified DereverenceTM to
		set hDClassName, and cChildren, GettChildTM to
		use hDClassName.
	debapi.h
		Minor changes due to hDClassName addition in the TM

29-Sep-1993 SanjayS
	debeval.c
		Don't force operand of ++ and -- to be treated as an lvalue in
		EvalPostIncDec and EvalPreIncDec.

28-Sep-1993 SanjayS
	debbind.c
		Free the correct memory handles in BinaryOperator, even if a
		realloc happens.

27-Sep-1993 MarkBro
	debfmt.c:
		Added fix for ensuring that the types table checked for a type
		matches the hmod of the symbol.

24-Sep-1993 Georgiop
	debfmt.c
		Use the mod for the type we back from SHModFromCxt when
		formatting a pointer to function value.
24-Sep-1993 Georgiop
	debfmt.c
		Dolphin #1389: Changed Format() to set the proper hmod in the
		evaluation node.

25-Aug-1993 Georgiop
	debapi.c
	dogfood.mak
	eent.mak
	makefile.mak
		Replaced LANG and LANGID macros with EELANG and EELANGID respectively

20-Aug-1992 SanjayS
	debsup.c
	debwalk.c
	debexpr.h
		Checked in changes made by Georgiop to fix problem where any TM
		created using EEGetTMFromHSYM would return a bogus string when

		EEGetExprFromTM was called on it.  See log made by Georgios below...
		N.B. Georgios was in Greece so his changes.txt was checked in later...

19-Aug-1993 Georgiop
	debsup.c
	debwalk.c
	debexpr.h
		Changed GetExpr to replace HSYMs found in the expression string
		with the corresponding symbol names.
	dogfood.mak
		Added /MD /W3 to the compiler command line

12-Aug-1993 MikeMo
	debtyper.c
		Fix a bug I introduced way back in February: when casting int to
		FOO*, where FOO* is not a primitive type, CastNode() was treating
		FOO* as a primitive type (e.g. doing CV_MODE on it and stuff).

10-Aug-1993 MikeMo
	debbind.c
	debparse.c
	debsup.c
	debsym.c
		Fix bugs in the use of MHMemReAlloc.  Some calls were assuming that
		the handle value could change, and others were assuming that it
		could not.  The former is correct: the handle value CAN change.

09-Aug-1993 GeorgioP
	debutil.c
		Modified erroneous (but harmless) code in GetUdtDefnTindex
		that generated compiler warning.

09-Aug-1993 GeorgioP
	Implemented "quick binding" for child TM's that represent member functions.
	Bind info are stored in the TM during the child TM creation and the bind
	phase is bypassed. Currently such TMs are not evaluatable. Affected files:
	debapi.c:
		Modified GetChildTM, added GetParmTM
	debeval.c:
		DoEval now fails gracefully if the TM has no eval stack.
	debexpr.h:
		Modified/added prototypes for GetChildTM, GetParmTM, ParseBind
	debsup.c:
		Modified GetChildTM, replaced SetClassiName with GetClassiChild

15-Jul-1993 GeorgioP
	debbind.c
	debeval.c
	debfmt.c
	debtyper.c
	debutil.c
		Fixes to avoid some compiler warnings

15-Jul-1993 GeorgioP
	debbind.c:
		One-line fix (Replaced '==' with '!=' on line 3023): AddrOf
		should not set EVAL_STATE(ST) to rvalue in case ST is a type
		(dolphin #1169)

15-Jul-1993 GeorgioP
    debparse.c:
	Added XFormExpr and GetToken for template support.
    debbind.c:
	GetId looks for matching <>'s in order to treat template class
	names as identifiers.
    debexpr.h:
	Added hExStrSav, strIndexSav, ExLenSav in exstate_t. These hold
	info about the original string in case it has been modified by
	XFormExpr.
    debwalk.c:
	GetExpr now uses the saved string if it exists.
    debsup.c:
	DupTM duplicates the saved string
    debapi.c:
	EEFreeTM frees the saved string

07-Jul-1993 DanS
    debfmt.c
	cuda:5647, display T_NOTYPE and default cases with 4 bytes
	for 32-bit addresses, 2 bytes for 16-bit.

02-Jul-1993 MikeMo
	debapi.c
		cuda#5595: EEGetTMFromHSYM now returns EENOMEMORY instead of
		EECATASTROPHIC if a memory allocation fails.  This allows the
		callstack to come up with an empty stack instead of crashing.

01-Jul-1993 DanS
    debfmt.c
    debbind.c
	cuda:5546, transfer all addr bits from ptr to sym in ref case.
	initialize type in AddrOf case for &(int &bar) case.

29-Jun-1993 DanS
    debutil.c
	cuda:5515, if type packing leaves dupes, we can end up with
	sym recs to have type indices that don't point into the correct
	type index "tree". the trick is to find a fix that won't break
	searching for ambiguous methods.  what we settled on was in fnCmp
	to a) if name matched exactly (case sensitive), b) the type that
	we were carrying around for the method was not the same as the one
	on the symbol record, then c) we would say they are equivalent if
	there were no overloads (overloadCount == 1).

22-Jun-1993 DanS
    debsym.c
	cuda:5345, don't let registers/line #'s be evaluated in an
	explicit class/struct scope.

\-------------------End Conflict-------------------/
16-Jun-1993 DanS
    debbind.c
	cuda:5171 (and more!) propagate an explicit context from the
	left bind tree to the right during BindDot and BindPointsTo.
	this handles the case where we have {,,foo.dll}fooStruct and
	we want to expand it.  trouble is that the expansion results
	in an expr like this: ({,,foo.dll}fooStruct).fooMember.  the
	context is bound to the right side and not applied to the
	right hand side for the type lookups.

	an alternative fix would be to always carry a context around
	on every symbol node or to elevate the context precedence so
	that we end up with {,,foo.dll}(fooStruct).fooMember instead.

14-Jun-1993 DanS
    debbind.c
    debeval.c
	cuda:5124, vtshape tables not dumped correctly.
	debeval.c wasn't passing on all of the address from ptr to sym
	in VFuncAddress.
	debbind.c had some unused code removed.

28-May-1993 DanS
    debsym.c
	cuda:4886, add case S_LTHREAD32 to SymToNode so that we can
	look at static (local and module) TLS variables.

25-May-1993 DanS
    debsrch.c
	cuda:4798, use all 4 bytes of 32-bit regs when loading symval
    debutil.c
	cuda:4799, xfer all data pertaining to register values from pv
	to nv in SetNodeType for any case that is a register.

25-May-1993 MikeMo
	debfmt.c
		cuda#4793: EVAL_IS_NPTR32 and EVAL_IS_FPTR32 return FALSE for
		arrays, which was causing trouble for code which assumed they
		would return TRUE for arrays as well as pointers.

06-May-1993 RicoM
    debsym.c
	CUDA #4305: SetValue() now also avoids referencing "this" when
	accessing a BASE class name that was found in the context of
	a member function.  It used to turn FOO::bar into this->FOO::bar
	if you were in a member function of a class derived from FOO.

06-May-1993 RicoM
    debsym.c
	CUDA #4030: SetValue() now avoids referencing "this" when accessing
	static data members from a class that was implicitly in scope.
	This lets us refer to a static data member from within a static
	member function without having to qualify with the class name.
	The old way we'd be told that we could find "this" even though
	we didn't really need it.

05-May-1993 RicoM
    debutil.c
	CUDA #4067: SetNodeType() needs to preserve the EVAL_IS_LABEL
	bit or else labels get treated as data and can't be fixed up
	correctly

04-May-1993 DanS
    ldouble.c
    ldouble.h
    debfmt.c
    deblex.c
    eent.mak
	cuda:2119, low budget float10 support using CRT internals.

    debfmt.c
    deblex.c
    debeval.c
    debsrch.c
    debsup.c
    debsym.c
    debtyper.c
    debutil.c
    debwalk.c
	change near to NEAR all over the place because new compiler removed
	near do nothing keyword

03-May-1993 RicoM
    debfmt.c
	CUDA #4028: FormatLong() renamed to ExtendToLong() which more
	accurately reflects what it does.  Bug #4028 says that
	(unsigned char)255,d formats as '-1'.  The problem here is
	that in the ,d case ExtendToLong() always did a sign-extend even
	when converting from an unsigned type.  There is now a check
	to see if we are converting from a unsigned type, or if we need
	an unsigned result.  In either of these two cases no-sign extension
	is performed.

30-Apr-1993 RicoM
    debeval.c
	CUDA #4044: VerifyFormat needed to check the return code of
	its CastNode() calls to prevent trying to format a float as
	memory for instance (i.e. float f; ?f,m)

30-Apr-1993 RicoM
    debeval.c
	CUDA #4020: EvalFunction stack handling changed so that
	any temporary objects that were pushed on the stack (e.g.
	string literals are not removed until the whole expression
	has been evaluated.  This handles nested/multiple functions

29-Apr-1993 RicoM
    debsrch.c
	CUDA #2962: LoadAddress wasn't always using pExState->frame.DS
	for the segment value (sometimes it used just 0)

29-Apr-1993 MikeMo
    debfmt.c
	cuda#4043: The IDE displays chars with null value as '\x00',
	but CV displays them as ''.

29-Apr-1993 RicoM
    debeval.c
	EvalFunction and VFuncAddress2 both preserve the EMI and MODE
	parts of the address when they copy the results of EvalUtil()
	into the address part of the expression.  This allows us to
	correctly fixup/unfixup vtable addresses...  EvalFunction
	was also modified to correctly compute the logical address
	of a near virtual function without consulting CS (which doesn't
	work anyway)  CUDA #3949 is now completely fixed

28-Apr-1993 RicoM
    debeval.c
	EvalFunction doesn't try to restore the registers if it fails
	before they've been saved anymore...  This makes CUDA #3949
	not fault anymore, but we still can't call the function...

27-Apr-1993 MikeMo
    debsup.c
	Revision of the change mentioned below on 21-Apr: this change
	is in Cuda after all.  EEInfoFromTM won't automatically
	dereference pointers.

26-Apr-1993 RicoM
    debbind.c
    debeval.c
	CUDA #3743, use 32 bit types in bind/eval of 'by', 'wo', and 'dw' ops

23-Apr-1993 RicoM
    debeval.c
	Push function addresses on the stack in EvalFunction in the prototype
	case.  CUDA #3673

21-Apr-1993 MikeMo
    debsup.c
	Under Dolphin, EEInfoFromTM won't automatically dereference
	pointers.

09-Apr-1993 SanjayS
    deblex.c
	Added additional overflow check in ParseIntConst.

08-Apr-1993 SanjayS
    debeval.c
    debbind.c
	EvalSegOp and BindSegOp were disallowing T_INT4's. Fixed this
	as integral constants are always typed INT in 32-bits. Added
	check in EvalSegOp to make sure that the value of the segment
	portion is in range.

08-Apr-1993 MikeMo
    debexpr.h
    debbind.c
    debeval.c
    debfmt.c
    deblex.c
    debparse.c
    debsrch.c
    debtyper.c
	* exstate_t.state now contains an 'f32bit' flag, which indicates
	  whether this expression has 32-bit ints and pointer offsets.  This
	  should usually be used instead of the 'fWin32' global (now removed)
	  or the 'CVTargetInfo' function.

	* Various expressions that used to return type T_SHORT or T_USHORT for
	  C boolean operations like "a != b" and "!x", have been changed to
	  return T_INT2 or T_INT4 as appropriate.

	* T_RCHAR is now treated as a signed char as it should, rather than
	  unsigned.  (There is still one case we don't handle correctly, which
	  is too hard to fix: if someone compiles a C++ program with -J, we'll
	  still be treating chars as signed.  But this is a rare case.)

	* In hex display, chars are always displayed with two hex bytes even
	  if variable's value is negative; shorts always with four hex bytes.

	* Fixed value display of things like "sc,d" and "sc,u" where sc is a
	  signed char.

	* When lexing a constant for 32-bit, return T_LONG or T_ULONG if user
	  explicitly appended 'u' and/or 'l'.

	* Fixed rules for "integral promotion."

	* Fixed rules for the "usual arithmetic conversions."

06-Apr-1993 MikeMo
    debtyper.c
	Cuda#2616: "eip == label" did not work, because the code was
	attempting to cast the label from a 32-bit address to a 16-bit
	address.

30-Mar-1993 Brucejo
    debutil.c
    debsup.c
	Fix 16 <--> 32 bit ptr casting problem.  This caused the array
	expansion to fail.  The macro N_EVAL_PTRTYPE was checking more
	than just the type, the tests were changed to check just the type.
	Also, examine return from SetNodeType and assert on false for the
	array element counting.  This should help catch this kind of bug
	easier.

04-Mar-1993 MikeMo
    fmtstr.h
    debfmt.c
	Clean up code which maps a primitive type to a string.  In
	the process, fix cuda#2558 (vars of type "int *" not displaying
	their type correctly).

29-Feb-1993 DavidGra
    debsup.c
    debsym.c
    debutil.c
	Added S_[L|G]Thread cases where appropriate in order to support
	thread local storage.


08-Feb-1993 MikeMo
    debbind.c
    debeval.c
    debfmt.c
    debtyper.c
    debutil.c
    fmtstr.h
	Fixes for 32-bit pointers.  Main fix was in CastNode() in
	debtyper.c: we now correctly cast between 32-bit pointers to
	different types.  Other small fixes.

01-Feb-1993 MikeMo
    makefile
	Added -Zp4 to NT builds.  We haven't tracked down why this is
	needed, but it is.

30-Jan-1993 MikeMo
    debfmt.c
    debsup.c
    shfunc.h
    makefile
	Added support for the debugger to pass a flag which indicates
	whether it wants hex constants to be in upper case or lower
	case.  (Flag is in eeapi.h.)  This option is ON by default
	when building the NT DLL, OFF for all others.

	Also cleaned up some M68K code that had far more ifdefs than
	it needed to.

09-Jan-1993 Georgiop
    debbind.c:
    debeval.c:
    Modified BindContext and EvalContext to treat bp-relative l-values
    without converting them to r-values if the explicit context is the
    same as the current context (workaround for caviar #6902).

07-Jan-1993 Georgiop
    debfmt.c:
    Extended EvalMem functionality to handle word, dword and ascii
    display (i.e., now the following specifiers are accepted:
    m, mb, mw, md, ma. The default (m) is byte display.

05-Jan-1993 Georgiop
    debexpr.h
    Made SearchBasePtrBase global

    debutil.c
    Now SetNodeType calls SearchBasePtrBase in order to
    find the base symbol of a based pointer.

    debsym.c
    Disabled calling of SearchBasePtrBase from Searchsym.

05-Jan-1993 Georgiop
    debfmt.c
    Implemented 'm' format specifier for memory dumps. "addr,m" displays
    16 bytes in hex and their ascii equivalents if printable.

29-Dec-1992 Georgiop
    debsrch.c
    Changed LoadSymVal to load additional bytes if a bitfield crosses
    a byte boundary (caviar #6610). Also fixed an off-by-one error
    in the index used for accessing BitsToBytes[]. (Related bugs: 3120,
    3107)

    debeval.c
    SBitField was polluting neighboring bits during a bitfield assignment.
    Now LoadSymVal is called after the type of the node has been set to
    the underlying type of the bitfield.

23-Dec-1992 Georgiop
    debapi.c
    Defined two new functions FpMask and FpRestore in order to mask
    and restore floating point exceptions under WINDOWS3. This is
    now used by EEParse, EEvaluateTM and EEGetValueFromTM. (affects
    caviar #6765 and #6372)

22-Dec-1992 Georgiop
    debutil.c
    Restored old behavior of SetNodeType regarding based pointers
    and added an extra symbol search. If the symbol on which the
    pointer is based is bp-relative we get the base information from the
    copy of symbol record in the type section, otherwise we call SearchSym
    and do a module/global search. This fix should permit dereferencing
    based pointers that are members of a struct, or double dereferencing
    of a pointer to a based pointer. (caviar #6413)

18-Dec-1992 Georgiop
    debapi.c:
    Mask fp exceptions before calling DoEval --cav. #6372

17-Dec-1992 Georgiop
    debexpr.c:
    Fixed return value in GetChildTM prototype (EESTATUS instead of bool_t).

    debsup.c:
    GetChildTM now returns EEGENERAL if the parent TM is not expandable.
    This should fix caviar #6381

14-Dec-1992 Georgiop
    debexpr.h:
    Temporary change: added CVFTargetInfo() until it is implemented by
    the kernel.

    debbind.c:
    Modified ParseType to handle 32-bit ptrs (cuda #960)

    deblex.c:
    Modified ParseString for 32bit support (cuda #908)

    debtyper.c:
    Modified CvtPtr, CastNode, CvtSizeDown, NormalizeBase for 32bit support
    (cuda #857)

    debsym.c:
    SetAmbiant now supports NEAR32  (cuda #960)

05-Dec-1992 Georgiop
    debapi.c:
    EEGetChildTM now initializes pExState puts any error messages
    to the parent TM. If it fails it frees the child TM. Caviar 5387

    debfmt.c:
    Removed "NYI" message. FormatClass now prints "{...}" in all cases.

    debsym.c:
    MatchArgs: If the argument is a type, require an exact match (i.e.,
    identical type indices). In general when matching args we
    ignore certain modifiers (e.g., const). In this case we want to
    be very strict, since we are probably dumping a class and we want
    to let all overloaded methods bind properly. Otherwise a "symbol
    is ambiguous" message may appear instead of a function name when
    dumping a class.


03-Dec-1992 MikeMo
    deblexer.asm
	Fixed 32-bit version to have the correct size for the
	elements of the LexTable: 6 bytes per element rather
	than 4.  Now lexing works.
    deblex.c
    deblexer.asm
	Changed ParseOp from pascal to stdcall.  (Actually in deblex.c
	it's defined PASCAL, which is stdcall.)

01-Dec-1992 Georgiop
    errors.h:
    Added ERR_NOSTACKFRAME: "variable needs stack frame"
    debsym.c:
    Changed SearchSym to stop the search for a symbol that has
    been found and has been rejected because of the fact that we
    are in the prolog of a function. Instead of searching the
    parent scope we now set an error message and return HR_error.
    this fix should also handle the case ?{foo}i where foo is
    a function that has not been invoked yet and i is a local
    variable in the scope of foo. In that case BindContext will
    create a context based on the address of foo, and things
    will go on as if we were in the prolog of foo. --caviar #5898

25-Nov-1992 Georgiop
    Caviar #5563: error gracefully when a FORTRAN complex type is
    encountered. Affected files:
    errors.h:
    Added new message "type not supported"
    debutil.c:
    Error detection in UpdateMem, TypeSizePrim
    debtyper.c:
    Error detection in IntegralPromote, CheckType, CastNode
    Also changed CastNode to allow casting of an enum to an int.
    debsrch.c:
    Error detection in LoadSymVal
    debbind.c:
    Error detection in BindByteOps, Function


21-Nov-1992 Georgiop
    debeval.c:
    Added code in PlusMinus to compute the address of a huge array
    element. --caviar 2120

    Disabled function evaluation for functions that return large
    classes or structs (this is because of differences in calling
    conventions that do not appear in codeview information.)

    Changed EvalDot to call EvalStruct instead of EvalElem if the
    Right child is a reference. This should correct the problem
    of miscalculating the return value when evaluating a function
    that returns a reference to a class. (Changes related to caviar
    #5425.)

18-Nov-1992 Georgiop
    debparse.c:
    Added a special case in FParseExpr in order to parse nested
    function calls of the form "fun1(arg1,fun2(arg2))" --cav. #3678

    debsym.c:
    The test in SearchSym that restricts the function search if an
    explicit context is given was moved one level up (from SYM_EXE
    to SYM_GLOBAL) because of the changes in the symbol handler.
    In addition this test has been enhanced in order to terminate
    the search only if a symbol has already been found. This should
    fix the problem described in caviar #5321 (empty error message).

    CheckDupAmb uses function addresses again in order to determine
    if a function is already on the list. A new ADDR field has been
    added to search_t struct for this reason. --caviar #5564

    Merged change from .420 tree: added explicit casts to (char far *)
    for the args of _fstrncmp(). This should not affect the generated
    code, since we compile large model anyway. --#5575

    debexpr.h:
    Added ADDR field in search_t. Removed the old address_t field, as
    it was not being used.

    debfmt.c:
    Added missing quote (') in _fmt_char_zr --caviar #5513

12-Nov-1992 Georgiop
    debsym.c:
    Changed MatchArgs to tolerate duplicate functions. pName->best is not
    updated if the current function is the same as the best one so far.
    Changed CheckDupAmb to compare function names and types instead of
    addresses. (It was broken because it was accessing a volatile
    evaluation node) -- Caviar #4791

11-Nov-1992 Georgiop
    debsym.c:
    DebLoadConst does not change the type of an enum to the underlying
    type any more. This helps proper argument matching. (caviar
    #3716, 3718).
    debfmt.c:
    Format(): In order to display an enum constant as an int, cast it
    to the underlying type before displaying it.

05-Nov-1992 Georgiop
    debeval.c:
    EvalAddrOf now sets an error message instead of asserting
    if it sees anything but an l-value (caviar #5191)

    debsym.c:
    Fixed coding error: '==' was being used instead of '=' in some
    assignements related to pExState->err_num. (caviar #4517)

28-Oct-1992 MikeMo
    deblexer.asm
	For NT, changed from C to PASCAL.
    deblex.c
	Under NT Beta 1, PASCAL is defined to __stdcall, but we want
	to declare ParseOp (which is defined in deblexer.asm) as actually
	being pascal, so I changed the prototype from using PASCAL to
	using _pascal.
    shfunc.h
    debapi.c
	Use EXPCALL calling convention for exported functions.
    eent.mak
	No longer need to invoke CVTOMF.

27-Oct-1992 Georgiop
    debsym.c:
    Changed SearchSym to handle failures of PushStack. This should
    prevent a GP fault in low memory conditions (caviar 4075)
    debutil.c:
    Added check in GrowStack to avoid unlimited growth of stack.
    Removed corresponding check from PushStack. (caviar 4075)
    debfmt.c:
    Increased the size of tempbuf in VerifyFormat so that FLT_MAX
    can be printed using %f format. Doubles and Long Doubles that
    are too large to be printed in %f format, are now printed in %e
    format. (caviar #4049, also related to #933).

20-Oct-1992 MarkBro
    debapi.c:
	Changed order of checking in EEGetExtendedInfo.  It was checking
	for ETIPOINTER before ETIARRAY.  The problem was that arrays
	are always pointers, but pointers aren't necessisarily arrays.  So
	now, check to see if the htm is an array before a pointer.

16-Oct-1992 Georgiop
    debeval.c:
	Added VFuncAddress2 function in order to compute the address of
	a virtual function if the offset in the vtable is given
	(VFuncAddress does a similar thing, but uses the function index,
	not the offset). This should solve caviar #4094. (Also inlined
	these functions as they were called only once.)

    debsym.c
	Modified MatchArgs in order to allow matching a far pointer with
	an array that has the same underlying type. Based pointer matching
	is not supported. (caviar #3713)
	Also changed definitions of vtabind (UOFFSET instead of ushort) and
	made the EE skip 4 bytes instead of 2 when accessing the vtab offset
	in an LF_MLIST record. (caviar #4342)


16-Oct-1992 MikeMo
    cvinfo.h
    cvtypes.h
    eeapi.h
    shapi.h
    types.h
	Deleted, moved to the CVINC.420 project.

    makefile
    *.mak
	Modified to support the new CVINC.420 project.

15-Oct-1992 Georgiop
    debsym.c:
	Modified MatchArgs to handle actual arguments that are references.
	This should fix the third function call mentioned in caviar #4182.
	Also changed two error messages to ERR_AMBIGUOUS instead of
	ERR_ARGLIST.

14-Oct-1992 Georgiop
    debbind.c:
	Changed the way VerifyFormat treated floating point format
	specifiers: Now g,e,f can be used with any floating point
	type. All prefixes except h are accepted, but are actually
	ignored: the EE uses the right prefix based on the size of
	the corresponding size. (caviar #933, #4058)

13-Oct-1992 GlennN
    debsym.c:
	 bug 4380. SearchRoot() changed to ignore returns
	 of NOTYPE when it encounters a nested unnamed ENUM
	 type. By doing this searches of classes with
	 unnamed enums will not abort prematurely. After
	 review from SteveSm We put a check around the
	 break, we only do this if we are binding a
	 breakpoint so that we minimize the potential
	 impact.

11-Oct-1992 MikeMo
    debapi.c
    debbind.c
    deberr.c
    debsup.c
    debsym.c
    debexpr.h
    errors.h:
	Added hErrStr to the exstate_t structure to save additional info
	about an error, so that if a "symbol not found" error occurs we
	can tell the user what symbol wasn't found.
    makefile:
	Changed NT build to use -Z7 instead of -Zi, and to not use -f
	since cl3232 doesn't recognize it.

10-Oct-1992 MikeMo
    *.mak makefile:
	Cleaned up some of the make stuff, to reduce needless repetition.
    types.h:
	SEGBASED(x) maps to _based(x) for 16-bit, nothing for 32-bit
	INLINE      maps to __inline (the EE uses this)
	_fstrcmp    maps to strcmp on NT
    debapi.c:
	Removed eedllinit(); link sets entry point to _CRT_INIT directly.
    debbind.c deberr.c debfmt.c debeval.c deblex.c debparse.c debsym.c:
	Use SEGBASED instead of _based.
    deberr.c debfmt.c:
	Fix lots of bogus based/unbased warnings.
    *.*
	Lots of fixes to get it working under NT.

07-Oct-1992 Georgiop
    debbind.c:
	BindBScope now reorganizes the tree if its left child foo has
	been rewritten as "this->foo". Reorganizing is necessary to
	restore operator priority (the bscope operator has higher priority
	than '->'). This should fix caviar #932, 2571.

06-Oct-1992 Georgiop
    debparse.c:
	Added a static table with the primitive type names. This table
	is scanned when an expression like "(token) op ..." is parsed,
	where op is an operator that has a binary and a unary form (e.g,
	+, -, *, &). If "token" is a primitive type name we know that
	the expression is a type cast and the operator is unary
	(caviar #3836)


05-Oct-1992 Georgiop
   debeval.c
       Added a new function (IsStringLiteral) that gets called by
       EvalFunction before the invokation of PushString.
       PushString is not called unless there is actually a quoted
       string in the expression that needs to be pushed on the
       user's stack. (this should solve caviar #4152)


05-Oct-1992 v-natjm
   makefile:    Modified these two files to handle the new Visual
   eew0cxx.mak: C++ DLL names

04-Oct-1992 Georgiop
    debsup.c
    'format' pointer in GetChildTM was incorrectly typed as static.

04-Oct-1992 Georgiop
    debeval.c
    Fix for caviar #932,4092,4099: EvalFunction fakes a stack entry
    before evaluating Lchild (if the Lchild node is OP_bscope).

04-Oct-1992 MikeMo
    debapi.c
	Fixed EEGetAccessFromTM: (1) it failed to handle a number of
	error cases; (2) it was allocating a much bigger buffer than
	it needed.  Also commented the little bugger.

	Changed two uses of "C_only" to "C_ONLY".

02-Oct-1992 JimSa
    debbind.c
	Need to fail for _segment in BuildType() when debugging 32 bit .exe.
	(Cuda #962)
    debsym.c
	Needed to set addr.flat32 bit for S_*32 symbols in SymToNode().
	This was only bein done for M68K but needed to be done for all
	32 bit expression evaluation. (Cuda #1030)

01-Oct-1992 Georgiop
    debsup.c:
    GetChildTM uses radix of parent TM for generating array indices (caviar
    #3910, see also #872)

30-Sep-1992 GlennN
    debapi.c
    eeapi.h
    shfunc.h
	added parameter fForceBind to pass to EEBind from this call.

30-Sep-1992 Georgiop
    debeval.c:
    Changed Relational() and Arith() for special handling of near null
    pointer comparison (caviar #3699, #3654)
    DoEval() now allocates EStack space if hEStack is Null (caviar #3804)
    EvalFunction() was not handling correctly references to integers
    constants (caviar #3712)

    debbind.c:
    PushCArgs, PushPArgs, PushFargs: if the argument is a reference use
    SetNodeType for the underlying type in order to remove modifiers.
    (caviar #3712)

    errors.h:
    Changed "implicit conversion required" message to "argument type not
    supported" (caviar #3843)

29-Sep-1992 v-natjm
   eew0cxx.mak:
      Added /NOE to link command line to avoid the linker error on multiple
      symbols for WEP (Build with C800)

21-Sep-1992 Georgiop
    debsym.c:
    SkipModifiers now handles primitive types

18-Sep-1992 DanS
    eentcxx.mak
    remove ntmain.obj, -base switch, use libcmt.lib
    debapi.c
    add call to _CRT_INIT from eedllinit
    coffbase.txt
    ntmain.c

    removed from project
17-Sep-1992 Georgiop
    debexpr.h
    Added overloadCount field in search_t.
    debsym.c:
    Changed MatchArgs, MatchFunction and MatchMethod to require exact
    type matches for overloaded function calls.
    debbind.c:
    Disabled error checking for string literals.
    BindCast now propagates type info to the OP_CAST node. This is necessary
    for calling functions with cast string literals as arguments.
    debeval.c:
    Changed the conditions under which PushString is called by EvalFunction
    debutil.c:
    SetNodeType now sets PTR_UTYPE for primitive pointer types
    (changes related to caviar #919)

10-Sep-1992 Georgiop
    debsym.c:
    Changed SetBPValue to save the offset of a class member (caviar #1756)

10-Sep-1992 DanS
    debeval.c:
    far -> FAR (portability)
    deblexer.asm:
    rewrote ParseOp for 32-bit i386, used HOST32 macro to key 32-bit code.
    debsym.c:
    far -> FAR (portability)
    debutil.c:
    far -> FAR (portability)
    makefile:
    added eentcxx
    opprec.c:
    bug on using pointer from strpbrk w/o checking for NULL
    eentcxx.mak:
    added for NT build of EE
    coffbase.txt:
    HACK! copied from NT for basing DLLs in NT.

4-Sept-1992 Rossg
    cvinfo.h
    Update to cvinfo with LF_VARSTRING numeric leaf

03-Sep-1992 Georgiop
    debutil.c:
    TypeDefSize now handles T_CHAR, T_UCHAR, T_INT2, T_UINT2
    T_INT4, T_UINT4 when dealing with bitfields.

01-Sep-1992 Georgiop
    debbind.c:
    debeval.c:
    Changed T_PCHAR to T_PRCHAR so that string literals can be
    passed as arguments to functions. (Caviar #3127 -- this was
    a regression caused by the changes on 16-Jul-92)

28-Aug-1992 Georgiop
    debbind.c:
    Bug fix in ParseType: restore original start and length of type
    string.
    debtyper.d:
    Do not allow casting of an object to a pointer.

28-Aug-1992 Georgiop
    debapi.c:
    Do not print "public " when displaying access type info.

25-Aug-1992 Georgiop
    eeapi.h:
    debapi.h:
    Added extra argument in EEGetAccessFromTM, so that one can choose
    between the full text and the abbreviated version

24-Aug-1992 Georgiop
    debeval.c:
    Changed EvalFunction to call EvalUtil in case the formal argument
    type is a reference to a class. (caviar #914)
    debsym.c:
    MatchArgs now allows references. Both changes are for caviar #914

20-Aug-1992 Georgiop
    debsym.c:
    Removed a line from MatchFunction that accessed pSymClass. This
    line had been added (incorrectly) in order to support
    propagation of access type info, and could produce a GPfault.

19-Aug-1992 MarkBro
    debapi.c:
	Added missing entries in the EXF struct in EEInitializeExpr.

18-Aug-1992 MarkBro
    dosdll.asm:
	Added missing entries in the dosx exe header.

16-Aug-1992 Georgiop
    debapi.c:
    Comment out DASSERT in EEGetAccessFromTM (conversion functions
    did not bind ok, so EEGetAccessFromTM was hitting the assersion)
    debeval.c:
    restore access flags after evaluation
    debsym.c:
    support propagation of access type information

11-Aug-1992 Georgiop
    debapi.c:
    ParseBind now uses the radix stored in the TM. This is a
    fix for Caviar #872

23-Jul-1992 Georgiop
    debeval.c:
    EvalBScope faked up a stack entry if CLASS_GLOBALTYPE(pv) was
    true, and the stack was empty. However, in cases like ?C::foo(5)
    CLASS_GLOBALTYPE may be true, and the stack is not empty (i.e., it
    contains the argument to the function), yet a separate entry is
    required on the stack. The change was to enable faking up a stack
    entry in this case.

23-Jul-1992 Georgiop
    debsym.c:
    Disabled the code in SearchQualName that tried a symbol even if
    there was not exact type match when binding breakpoints.
    According to RossG this was a hack for an incomplete pack
    problem that should have been fixed by now. Disabling this
    code resolves Caviar #1081.

23-Jul-1992 Georgiop
    debapi.c:
    accessstr = {"   ", "private ", "protected ", "public "}
    (Changed uppercase 'P' to lowercase after a request from
    Sequoia people)

21-Jul-1992 Georgiop
    debbind.c:
    Fixed ParseType to retain type information if the type is an alias
    to a primitive type. [Caviar # 2271]

17-Jul-1992 Georgiop
    debfmt.c:
    Fixed bug in FormatProc, in order to print the function return type
    if select is 0. This takes care of bugs 1142 and 2343. The kernel
    should also be modified to call EEGetTypeFromTM using a handle to
    a Name. That way EEGetTypeFromTM will insert the name in the proper
    place and a function name will be displayed as "void foo(void)",
    not "void (void) foo".

16-Jul-1992 Georgiop
    deblex.c:
    ParseString now sets type to T_RCHAR instead of T_CHAR. This
    takes care of bug 917, although the real problem (EE does not
    distinguish between explicit and implicit casts) remains.

16-Jul-1992 Georgiop
    errors.h:
    Added ERR_NOMODIFIABLELV (l-value is not modifiable)

    debbind.c:
    BindAssign now checks for non-modifiable l-values (arrays,
    functions, or constants) -- CAVIAR bug 952

09-Jul-1992 Georgiop
    debapi.c:
    eeapi.c:
    Added a new enumarator as a return value for EEIsExpandableTM:
    ETYPEPTR denotes that the TM is a type that is expandable as a
    pointer to a type. (The underlying type may or may not be
    expandable). This solves CAVIAR bug 927.