- reset file pointers to previous context after evaluating arguments
- in __fb_query_symbol__, __fb_eval__, __fb_arg_extract__, __fb_iif__,
- and after releasing lexer context
- add enum LEX_TKCTX_CONTEXT to specify the kind of lexer context
- LEX_TKCTX_CONTEXT_INIT    for initializing, first time only
- LEX_TKCTX_CONTEXT_INCLUDE for include file
- LEX_TKCTX_CONTEXT_EVAL    for preprocessor evaluation

- add support for zstring and wstring conversions to better support
  unicode source files
- builtin macros __fb_query_symbol__, __fb_arg_extract__, __fb_iif__
  internally will use hMacro_EvalZ() and hMacro_EvalW() to evaluate
  compile time expressions with appropriate string conversions
- add additional tests for sf.net # 994 - but using utf-8 and utf16le
  encoding

- reset file pointers after evaluating arguments in __fb_query_symbol__, __fb_eval__, __fb_arg_extract__, __fb_iif__
- Some fbc built in macros like create a separate lexer context to evaluate arguments at compile time.
- When the lexer context is no longer needed, the previous lexer context needs to be restored.
- fbc was failing to correctly restore file pointers after freeing the context
- this would result in the file pointers being not synchronized with the memory buffers and memory pointers
- this would happen when the macros and arguments to be expanded sit on either side of an 8Kb boundary
- 8KB is the hard coded buffer size for reading in chunks of source code

- ignore (parse) single line comments for macros invoked with
  optional parentheses instead of passing as an argument to macro

- Change overload to IMAGEINFO( byval as const any ptr, byref as long = 0, byref as long = 0, byref as long = 0, byref as long = 0, byref as any ptr = 0, byref as long = 0 ) as long

(cherry picked from commit b6d4d462)

This partially reverts commit 42a07d16.

- remove changelog message
- keep changed code but disable by preprocessor
- add notes for TODO and FIXME

- fbc: win32/win64: pass '-fno-ident' to gcc to help prevent
  identification strings from accumulating in the final binary

(cherry picked from commit e6bc5cc7)

Conflicts:
- changelog.txt: resolve by adding the message to fbc-1.10.1 history

- remove the 'FindWindow()' for console windows when running under Windows 10 or newer
- When running under windows terminal, the FindWindow() logic always fails and
  MULTIKEY() does not return true for any keys check
- instead perform an OS version check to determine if the console window should be searched

Add ticket numbers for changelog entries:

Related to commits:
- 467b8993
- 50620135

- decide on signed or unsigned division based on sign of the result type
  instead of exact match to result type
- previously LONGINT result uses signed division, and every other type
  would use unsigned division

- fbc: win32/win64: pass '-fno-ident' to gcc to help prevent
  identification strings from accumulating in the final binary

- Can't pass UDT member array to an overloaded sub
- commit #8ee7cd4e made internal handling on non-indexed arrays more strict but
  missed cases where they should in fact be allowed
- tests added

fixed:
- gas64: improvement of message when internal errors arise in gas64 emitter (SARG)
- gas64: handle priority for constructors/destructors (SARG)
- gas64: optimization 16 modified retrieval of registers and improved to handle specific cases (SARG)
- gas64: NEW UBYTE/BYTE[count] not handled (SARG)

- emscripten/fbc: don't cast to integer for array indexing
- possible regression of behaviour when fixing github #217
- since commit #44adf4f2

Purpose is to allow procedures in sources to be compiled as separate modules with
public linkage or as privately included with minimal overhead for changing desired
linkage.

- set default procedure linkage with
  '#pragma private [=value]' or
  '#pragma push( private [,value] )'

- #pragma private
- #pragma private = true
- #pragma private = false
- #pragma push( private )
- #pragma push( private, true )
- #pragma push( private, false )
- #pragma pop( private )

- add '#elseifdef SYMBOL' to test symbol defined
- add '#elseifndef SYMBOL' to test symbol not definde

Abbreviated conditionals:

- #ifdef A
- #elseifdef B
- #elseifndef C
-   ... etc
- #endif

Instead of:

- #if defined( A )
- #elseif defined( B )
- #elseif not defined( C )
-   ... etc
- #endif

- Default access to a member symbol is disabled by a duplicated symbol in implicit namespace
- regression of behaviour when fixing #948
- don't let local variables from the main scope override symbols in a type's namespace

- preserve the 8-bit boolean bitfield container size when shifting / masking bits

- fix sf.net # 984 - Incorrect syntax for a member array passed as argument to a procedure
- remove astRemoveNIDXARRAY(), deleting the AST_NODECLASS_NIDXARRAY node is handled differently now
- internally, be more strict about wrapping non-indexed arrays in a AST_NODECLASS_NIDXARRAY
- preserve non-indexed array marker (AST_NODECLASS_NIDXARRAY) while parsing and remove only
  when checking the AST for valid arguments / expressions

- add 'enum FB_WCHARCONV'
- replace 'env.wchar_doconv' with 'env.wcharconv as FB_WCHARCONV'
- when compiler is built with __FB_DEBUG__<>0 then allow wstring conversions
  even though the host wchar size and target wchar size are different
- purpose of this is to allow some building and testing with the test-suite
  on a win32 host while also not having deal with all unicode/wchar issues

It is a link-time flag and does nothing at compile-time. I think this has
changed since earlier Emscripten versions, due to Emscripten's new wasm backend.

- optimize SHL MOD INTDIV to use 32-bit operation when result will be converted to U|LONG
- convert 'clng(cint(a<32>) BOP<64> cint(b<32>))' to 'a<32> BOP<32> b<32>'

For example:
    dim as ulong a, b
    var x = a \ b         '' same as before, a & b promoted to integer, returns integer
    var y = culng(a \ b)  '' optimized to use 32-bit division, return ulong

- use this optimization for conversion to U|BYTE and U|SHORT results also
- enabled for x86_64 targets only

- code clean-up of previous commit; compare data type sizes instead of specific data types

- add recipe for building fbc using winlibs-gcc-11.2/3/4 toolchain

- update to build libffi 3.4.4
- this is needed to work with newer mingw-w64 based toolchains for gcc 11.x and up

- use 'char' type instead of 'uint8' for zstring to avoid warnings in newer gcc 11+
- this avoids gcc warnings where fbc's crt headers are decalared in gcc but do
  not match the built-in prototypes for 'printf( const char *, ...)' and others

- add recipes for building fbc packages
- Mingw-w64 project gcc-11.2
- Mingw-w64 project gcc-12.2
- winlibs-11.3
- winlibs-11.4

- use 'll' instead of 'I64' if building for ucrt or
  mingw-w64's stdio implementation (for newer gcc
  versions and c std)

CURDIR works fully. I copied over rtlib/unix/sys_getcwd.c
EXEPATH always returns "/"

Previously, fb_hInit() caused in error in JSEvents.registerOrRemoveHandler
as 'window' was undefined.

JS strings are 16-bit (UCS2), which is where the confusion comes from.

In practice depending on exact use, using wstrings could seem to work but get
corrupted at the ends.

Previously, fbc did produce libraries (eg. fbcunit.a) but they produced an error
about needing to run ranlib when linked.

Finish the work in progress for handling NaN's

- gas backend: check for NaN's (parity checks) - 386, 486, 586, 686
- gas+SSE backend: check for NaN's (parity checks) - 386, 486, 586, 686
- llvm backend: not-tested
- gcc backend: already completed

- handle IR_EMITOPT_REL_DOINVERSE option in gas/gas+SSE/gas64 backends

The four scenarios passed to backend as part of the float comparison op:
- result := (a rel-op b)
- result := !(a rel-op b)
- if (a op b) then goto {falselabel}
- if !(a op b) then goto {falselabel}

test-suite: tests for float comparisons

Backends tested:
- gas: 386, 486, 586, 686, and each with SSE
- gcc: 386, 486, 586, 686, and each with SSE
- gas64: x86_64
- gcc64: x86_64
- tested all above combinations with '-fpmode fast' (Extra code
  added by the NaN checks is removed with '-fpmode fast')

- fbc: internal add AST_OP_BOOLNOT logical unary operator
- gcc backend: used internally by gcc backend for emitting the logical not '!' operator
- gas SSE backend: add operation to load Boolean from Float
- gas SSE/x86 backend: check parity (NaN) on Float to boolean conversion, extra check is removed
  when using '-fpmode fast' and enabled in '-fpmode precise' (the default).
- test-suite: add test for float to bool

The intent with this change is have cbool(float) behave consistently across all targets
- cbool( float ) is FALSE if and only if float = 0.
- For any other value (including NaN's) cbool(float) returns TRUE

This is a work in progress.  Updating the floating point comparisons
require changes in AST, IR, and all backends.  Seems useful to break
the overall changes in to a couple of commits.  This change is focused
on getting the new AST in place, though includes some llvm/gas64/gcc
backend updates and leaving the gas/gas+SSE for a later update.

- handle IR_EMITOPT_REL_DOINVERSE option in gcc/llvm/gas64 backends
- gas64 x86_64 backend : re-use astGetInverseLogOp()
- gcc backend all targets: use !(float op float) to implement reverse logic
- llvm backend: Add floating point comparison instructions,
  Note: currently no testing completed for NaN's in LLVM
- 'fpmode fast' will disable NaN checking - and will affect
  all backends.  Comparisons involving NaN's will be inaccurate
  however, all other comparisons of numbers (including infinities)
  should still be correct.
- idea is to let the backend optimize float comparisons where possible
- otherwise use safer code generation to get consistent results across
  branching, IIF, expressions, etc.

- allow '-fpmode fast' command line option for any target (not just SSE)
- this is to be used augment floating point handling where some accuracy
  is sacrificed for smaller and faster code generation
- this change only allows the option and does not affect any code generation
- previously an error message would be displayed if this option was
  with anything but SSE.

This commit adds some internal changes for tracking specific emit options in AST/IR
that can then be passed on to the backend / EMIT.  This will potentially provide a
generic way to extend AST options to the backend.

- add 'enum IR_EMITOPT' for options relating to EMIT options
- add 'AST_OPOPT.AST_OPOPT_DOINVERSE' for relational operators  to indicate that
  the AST expression needs to be inverted.  This allows AST to preserve the user's
  original expression and pass it on to the backend to solve.
- add 'IR_EMITOPT.IR_EMITOPT_REL_DOINVERSE' to indicate the option to the backend
- add 'EMIT_NODE.options' to track options emit options
- Update `EMIT_RELCB` to accept an options parameter
- For procedures handling relational operators in IR/EMIT, add parameters to the
  internal functions to accept 'options as IR_EMITOPT'.  This will then control
  how the backend handles code generation.