
Version 55
----------

 - the history file is started

 - get_name() and similar functions have two additional parameters:
        char *buf, size_t bufsize
   i.e. they return their answers in the specified buffer, not in
   a static buffer anymore

 - all processor modules should call invoke_callbacks() from their
   ph.notify function. See the sample file, reg.cpp.

 - sel_t type is 32-bit now (before it was defined as ushort)

 - plugin init/term() functions are called only once for each plugin.
   before they we being called for each menu entry

 - a plugin can return PLUGIN_KEEP from his init() function. In this
   case the plugin will stay in the memory. otherwise it is unloaded
   and loaded again only when the user invokes it.

 - notification callback functions are added

 - enumerations can have several constants with the same value
   to distinguish them we use "serial" numbers. Each constant in the
   enumeration has a serial number. Usually the serial number is equal
   to zero. If there are 2 constans with the same value, their serial
   numbers will differ. The serial numbers are allocated by the kernel
   when a symbolic constant is created. The kernel allocates the serial
   numbers starting from zero (0,1,2,3...) The maximum value for a serial
   number is 255.
   The interface to the enum oriented functions have been changed to
   include the serial numbers.

 - user-defined line prefixes are supported. see get_user_defined_line_prefix()
   and the sample plugin (vcsample).

 - The following functions have been renamed (the list is not exhaustive):
   itemEnd() -> get_item_end()
   getSize() -> get_item_size()
   doUnknown() -> do_unknown(), do_unknown_range()
   add_segm -> add_segm_ex
   set_refinfo -> set_refinfo_ex
   set_ti -> set_op_ti
   get_ti -> get_op_ti
   add_func -> add_func_ex
   apply_idasgn -> apply_idasgn_to
   autoMark -> auto_mark_range
   set_name -> set_name_ex
   set_name -> set_dummy_name
   doData -> do_data_ex
   QueueGet -> QueueGetType
   chsize -> echsize
   btoa -> btoa32, btoa64, btoa128
   b2a -> b2a32, b2a64
   atob -> atob32, atob64
   get_const -> get_const_by_name

 - The Visual C++ compiler can be used to write plugins.
   See the samples in the SDK.

 - A plugin can extend the processor module capabilites by intercepting
   custom_... callbacks. See the sample (procext).

 - CBuilder VCL components can be used in the plugins. See plugins\chart.


Version 56
----------

- is_spec_ea() doesn't pay attention to inf.s_specsegs

- AskUsingForm() with buttons was broken

- segment translation functions don't work with the segment selectors
  anymore. In order to break the old code using them, these functions
  are renamed: anyway the code should be revised.

- new processor_t.flag bits (please check your modules and add these
  bits if you need them):
#define PR_BINMEM     0x10000   // the processor module provides correct
                                // segmentation for binary files
                                // (i.e. it creates additional segments)
                                // The kernel will not ask the user
                                // to specify the RAM/ROM sizes
#define PR_SEGTRANS   0x20000   // the processor module supports
                                // the segment translation feature
                                // (it means it calculates the code
                                // addresses using the codeSeg() function)

- choose2() function has one more additional optional parameter

Version 4.18
------------

- PR_BINMEM is not used anymore


Version 4.20
------------

- ieee_realcvt() function is added

- AS2_STRINV is added

- choose_ioport_device() has 2 parameters

- better handling of inf.wide_high_byte_first

- OutValue(..., OOF_ADDR) accepts dt_code as dtyp.

- 0x80, 0x8000, 0x80000000 are considered as signed by default in OutValue()

Version 4.21
------------

- set_dummy_name has 2 parameters

Version 4.3
-----------

- search.hpp is included

Version 4.5
-----------

+ IDA environment variable is not required to build modules anymore
+ added comments about filling the op_t structure; fixed some typos in netnode.hpp
+ COLOR_INV is added
+ hidden plugins are supported: PLUGIN_HIDE flag is introduced
+ idaw choose() function respects the batch mode
+ negative buffer sizes are handled properly (str2user, user2str, pack_ds)
+ new function flag FUNC_BOTTOMBP. It means that the frame pointer is equal to the stack pointer in the function and it points to the bottom of the stack frame.
+ ph.flag PR_CHK_XREF: don't allow near xrefs between segments with different bases. This flag is used for IBM PC only.
+ read_ioport_device() function reports about configuration files with no devices
+ renamed FIXUP_PTR32->FIXUP_PTR16, FIXUP_PTR48->FIXUP_PTR32
+ the user-defined data supplied to linearray_t is documented in kernwin.hpp
+ up to 16 source files for plugins
+ setBreak() function is added
+ the processor extension callbacks are called for all instructions, not only when cmd.itype >= CUSTOM_CMD_ITYPE
+ find_ioport_bit() returns NULL is the bit name is NULL
+ rebase_program() is added. This function allows to shift the whole program in the memory. Since rebasing the program involves correcting the relocated bytes, the file loader takes part of the job. File loaders may have "move_segm" callback functions now.
+ now a good behaving procesor module handles the ph.move_segm event
+ numop2str(): output instruction operand with optional leading zeroes; is_lzero(),toggle_lzero() to modify the display of leading zeroes; inf.s_genflags introduced; atoa, b2a32, b2a64, b2_width function parameters has been changed
+ move_segm_start(), set_segm_start(), set_segm_end() may destroy the adjacent segment if necessary; ADDSEG_QUIET flas has been added
+ new type of segments: SEGM_DEBUG. Used in the debugger.
+ get_sourcefile() function prototype has been changed. Now it returns the range information.
+ hidden_area_t and functions to work with it are introduced
+ byteValue() function is renamed to _byteValue(); this function should not be used anymore if possible. The reason is that it works only with 8-bit processors and doesn't take into account possible debugger side-effects.
BUGFIX: qmakepath() could generate file names with several backslashes in them

Version 4.6
-----------

NOTE: IBM PC register encodings are changed! See intel.hpp for the details.

+ prototypes of some processor module provided functions are changed
+ moves.hpp, sistack.hpp are included into the sdk
+ dbg.hpp: the first version of debugger api
+ idd.hpp: debugger plugin api has been changed
+ all out.../Out... functions check the output buffer boundaries (some other functions too).
  It means that all old processor modules should be modified to support the new model:
    - u_line variable disappeared
    - init_output_buffer/term_output_buffer are added
    - tag_... functions have additional parameter
    - many other functions have additional parameter to check
      the output buffer size
  The modifications are not heavy - a normal module would require the replacement of 2 lines, the initialization of u_line and termination.
  The new model allows to detect the buffer overflows in the processor modules.
+ MD5 functions are available
+ calcexpr_long() accepts a pointer to uval_t as well to sval_t
+ can compile plugins with Visual C++ command line compiler (available either in Visual C++ 6.0/7.0 or as free with .NET framework SDK + Plateform SDK)
+ construct_macro() function is added
+ do_name_anyway() does not complain to the user about bad names anymore
+ fixed a bug in swap64() and swap128()
+ foreach_strmem() and get_struct_member() function prototypes are changed to handle the member field names
+ gcc can be used to compile IDA plugins, loaders, and modules
+ graphing functions are available in IDA API; added ctype.h to pro.h
+ import_node is available in IDA API
+ int128 type is added
+ is_call_insn() function and callback is introduced. The callback should be implemented by the processor modules with unusual call instructions (like PowerPC)
+ more floating point functions are exported
+ new function: reftype_t get_default_reftype(ea_t ea);
+ all out.../Out... functions check the output buffer boundaries
+ tag_skipcodes() function is added
+ zip compression handling functions are added to the API

Version 4.6 SP1
---------------

+ floating point conversion functions have additional parameter: the output buffer size
+ generate_disasm_line() has an additional parameter. Currently it can be used to force instruction decoding even if there is no instruction at the specified address
+ open_url() function is added
+ qfscanf() is added
+ debugger: functions to list, attach and detach processes are available for plugins
+ get_tilpath() accepts the output buffer as a parameter
+ added various keywords to display more complex message boxes (warning(), info(), etc)

Version 4.7
-----------

+ HIGH22 and LOW10 offset types are generalised to be VHIGH and VLOW. The processor module can specify the widths of these fixups in the ph.high_fixup_bits field. Currently they are used in the SPARC and HPPA processors.
+ NULL value may be passed as the tester function to the nexthat, prevthat functions. It means that any address satisfies the criterium.
+ PR_FULL_HIFXP is introduced. It means: VHIGH fixup type expects the operand value to be equal to the full address of the target, not only the high bits. Used for HPPA HIGH21 fixup types.
+ UI list functions (choose(), choose2(), ...) now support multiple selection => the delete callback prototype was changed accordingly (older plugins can simply return 'true' to remain compatible)
+ added possibility to pass command line options to plugins (get_plugin_options)
+ added set/get_idc_func_body() to avoid frequent recompilation of IDC functions
+ debugger: enable_XXX_trace() functions can now disable tracing but conserve trace-over breakpoints
+ gen_use_arg_types() is added
+ lread() function is added; this function should be used in the loaders instead of eread(). The lread() function verifies if the read is ok, if not, it informs the user about it and asks if he wants to continue. If the user does not want to continue, the loader_failure() function is called
+ regex_match() to match regular expressions is added
+ removed support for the watcom compiler
+ set_idc_func() to add/remove IDC functions written in C++
+ the kernel knows about macroinstructions (cmd.flags |= INSN_MACRO); fixup information for macroinstructions is handled in a special way: partial fixups are combined into one full fixup
+ AS2_BYTE1CHAR is added: for wide byte processors, one character per byte
+ added the FILE option to the AUTOHIDE keyword for message boxes, to save hidden message box results to IDAMSG.CFG
+ get_next/prev_member_idx() functions are added; guess_func_type() understands stacks growing up (not tested yet)

Version 4.8
-----------

The IDA API has been considerably modified. There were 2 reasons for this:

        - a vulnerability in the version 4.7 has been found. Because of this
        the API was revised with the security in mind and many functions
        got additional parameters (like the size of the output buffer).
        We banned some potentially vulnerable functions from IDA.
        Now you can not use sprintf, strcpy and similar functions in the
        modules. However, if you still want to use them, please define
        the USE_DANGEROUS_FUNCTIONS preprocessor symbol during compilation.

        The replacement functions behave slightly differently than their standard
        counterparts. Please read the descriptions. While the differences are
        minor, they are still important (for example, qstrncpy() always
        puts the terminating zero at the end and qstrncat() accepts the size
        of the output buffer as the third parameter).

        - IDA 4.8 can load and run remote files. For this, the file loaders
        were modified to work with "linput_t *" type instead of the "FILE *".
        The linput_t means 'loader input source'. The kernel opens a local
        or remote file and passes the handle to the file loader. The file loader
        does not know whether the file is local or remote, it just reads it
        as before using file manupilation functions provided by the kernel.
        So, qfseek should be replace by qlseek, qfread by qlread and so on.
        This modification is quite straightforward and should not pose any
        problems.


In general, we tried to modify the API in the way that breaks the compilation
process instead of silently modifying the function behaviour. So, the task of
porting the existing modules/loaders/plugins to the new version is quite
simple - try to compile the module. Get rid of the compilation errors by
adding new parameters to the function calls, or replacing the old function names
with the new names. If your compilation is successful, then the new module should
work with the new version.

Please see how the samples in the SDK were modified to have an idea how to modify
your custom modules.

We were also forced to modify the debugger API to support remote file loading, and
better catch network errors. Check the idd.hpp file for the details.

Also the method to obtain the pointer to the debugger description has
been changed. We don't use the run() method anymore.
Please see the sample linux debugger.

Below is the list of detailed modifications made to the API:

+ added functions to manipulate XML values in the internal XML tree
+ IDC functions can be called without a database. Built-in functions have a special flag EXTFUN_BASE telling that they must be called with an open database
+ SetFileExt, FileExt, hasExt functions are renamed as set_file_ext, make_file_ext, get_file_ext. Additional parameters to check buffer overflows are added
+ added a possibility to tell 'this is definitely not a thunk function' in the ph.is_jump_func notification for the processor modules; this leads to the slight API change which should not pose any problems with the existing modules
+ added more segment alignment codes
+ added netnode::end() and netnode::prev() functions
+ additional parameter for parse_type() - the name of the declared variable
+ bit masks are supported by get_name_value()
+ find_error(), find_notype(), tag_skipcode() functions are added
+ functions dealing with the signness and bitwise negation are revised; now the bitwise negation has a bit in the flags which allows to use it in the structure and enum definitions
+ gen_idc_file() is removed; use gen_file() instead
+ introduced 'stack variable scaling factor'. Used for processors which implicitly scale stack variable references, e.g. tms320c55
+ introduced the notion of 'ephemeral' segments: debugger segments which are not loader segments. The kernel does not analyze ephemeral segments.
+ introduced the notion of a 'loader' segment. All segments created by loaders have this attribute.
+ is_32bit_func() is replaced by get_func_bitness()
+ leading_zero_important() function has been added
+ only debugger segments can be created when the debugger is active. you can use change_segment_status() to convert debugger segments to regular segments afterwards.
+ ph.kernel_config_loaded callback has been added - for plugins who want to set their own config settings
+ removed the ph.align_size() callback from the type system callbacks.
+ shortened some field names in the debugger related structures (thread_id -> tid, etc)
+ string list is available for plugins. see strlist.hpp
+ new sdk function set_sreg_at_next_code(); ph.setsgr parameters are changed
+ introduced AS2_COLONSUF for assemblers with :xx suffixes in the address expressions
+ renamed 'analyse' -> 'analyze' in all function names
+ added auto_queue_empty callback
+ choose_ioport_device() has no default parameters (so we have a compilation error instead of wrong runtime behaviour)
+ get_member_ti() accepts NULL as the buffer
+ interface to the file loader has been changed. Instead of FILE *, now we use linput_t * (see functions in diskio.hpp). linput_t allows to work with local and remote files.
+ modified the interface of many functions to avoid buffer overflow vulnerability problems
+ new function: sanitize_file_name()
+ renamed cmangle_name() -> decorate_name()
+ splitSRarea() has been removed. Use splitSRarea1()
+ added func_t * to the display_flow_graph()
+ added qisdir() function
+ added qvector and qstring template classes to IDA API
+ added a comment about the necessity of using the invalidate_... functions from the debugger notification callbacks
+ method to obtain the debugger description has been changed: now the init() function of a debugger plugin must fill the 'dbg' global variable if it wants to stay in the memory; the run() method is not used for the debuggers anymore

Version 4.9
-----------

IDA API was improved for thread-safety but this work has not been finished.
Biggest problems arise from the functions returning pointers to internal IDA objects.
The kernel may destroy these objects any time and this will lead to a dangling pointer.
In the single-thread model things work well because the kernel takes special measures
to keep recently requested objects in the memory, but this approach can not work
in the multi-threaded environment. I guess we will need to add new 'multi-thread friendly'
versions of existing IDA APIs which do not return pointers but fill the specified buffer.
So in the future you might expect a new function like

bool get_func_ex(ea_t ea, func_t *buffer_to_fill);

instead of the current

func_t *get_func(ea_t ea);

But since we freeze IDA API today, the old functions will still be available. When (if)
we make a thread-safe version of the kernel, the old functions should not be used in the
multi-thread environments.

Many functions have also been modified to return ssize_t instead of char *.
There are 2 reasons for that: first, returning pointers is considered as a bad idea;
second, having ssize_t (the length of the answer or -1) makes it easy to fill the output
buffer and move the pointer. Typical code sequence might be

 char buf[MAXSTR];
 char *ptr = buf;
 char *end = buf + sizeof(buf);
 ssize_t len = ida_api_function(..., ptr, end-ptr);
 if ( len <= 0 )
   goto error;
 ptr += len;
 // and so on...

As usualy we tried to modify the functions in such a way that the compiler will report an
error for everything that requires your attention. Most of the time the required
modifications will be simple and standard. Please check the prototype and description of
the culprit function to deduce the required modifications.

Below is the list of detailed modifications made to the API:

+ SDK: IDA API has been frozen for future binary compatibility
+ SDK: add_regvar() modifies existing regvar definitions to define a new variable which overlaps the existing variables
+ SDK: add_sourcefile() function does not fail if there already was a source file defined at the specified range; in this case it will delete or modify the old definitions to make a hole big enough to hold the new file
+ SDK: add_stkvar2() function to add stack variables from plugins; tds plugin uses it
+ SDK: added a flag to allow the use of constructs not supported by the target assembler. It is INFFL_ALLASM bit in inf.s_genflags. The inf.use_allasm() function checks this bit
+ SDK: areacb_t::make_hole() function can be used to make holes in area definitions
+ SDK: AUTOHIDE NONE and similar keywords in the dialog box messages can appear not only in the format string but also in the final string as well
+ SDK: closing_comment() to get the comment closing sequence
+ SDK: debugger structures are aligned at 4 bytes to avoid problems on ARM processor
+ SDK: is_ret_insn() function and callback are added
+ SDK: make_visible_name() has additional argument - output buffer size
+ SDK: new function calc_bare_name() to get the smallest possible form of a name (try to undecorate and demangle)
+ SDK: new function parse_types(); IDC: new function ParseTypes()
+ SDK: new function: get_flags_ex(). The new function get_flags_novalue() built on top of get_flags_ex() does not return the MS_VAL and FF_IVL fields of the flags and therefore is much faster for remote debugging; it is not exported yet (todo later!) but the kernel uses it during the segment deletion which improves the speed
+ SDK: ph.guess_memory_model is replaced by ph.setup_til
+ SDK: rotate_left() function is exported
+ SDK: SetFlags, SetFlbits, ClrFlbits() functions do not modify the MS_VAL and FF_IVL flag fields; uFlag global variable does not contain these fields; the kernel updates the uFlag automatically (if the modified address is equal to cmd.ea), so there is no need to update uFlag manually in the emulator function of processor modules
+ SDK: setup_selector() allocates a new selector for values not fitting in 16 bits only for IBM PC. Other processors will use 32-bit segment bases
+ SDK: the processor name is stored in the debugger description structure to allow instant debugging for processors different from IBM PC
+ SDK: added new data type: 3byte; doData() function is removed because it can be replaced by do_data_ex()
+ SDK: btoa..() functions accept a buffer for the answer; atoa() function is obsolete and removed (use ea2str())
+ SDK: close_chooser() to close open non-modal list views
+ SDK: dbg->stopped_at_debug_event() does take require 'event' as the parameter;
+ SDK: elnum_t is replaced with int
+ SDK: func_does_return() is added; it is better to use this function instead of examining FUNC_NORET flag directly
+ SDK: get_name_expr() accepts a buffer for the answer and returns the answer length; new function: out_name_expr() is easier to use in the processor modules; append_disp() is replaced with print_disp() which does not append but simply uses the given buffer and returns the answer length
+ SDK: IDC compile/run functions return the error message in the specified buffer (before it was a in a static storage); the return value is a bool meaning success of the operation
+ SDK: interface to netnode functions returning big objects has been changed. Now these functions put the results into the specified bugger. This is a big change in IDA API leading to the modifications of many functions. This change is a small step in the direction of multithreaded model.
+ SDK: is_alloca_probe notification code has been added
+ SDK: lexical analyzer is thread-safe and can be called from several threads for different inputs
+ SDK: new function qerrstr() and qerrcode()
+ SDK: nexthat, prevthat functions accept 'user_data' parameter for thread-safe handling
+ SDK: now the memory config and contents are not automatically refreshed at each debug notification; the plugin must call invalidate_dbgmem_config() and/or invalidate_dbgmem_contents() to get the current view
+ SDK: PLUGIN_PROC and PLUGIN_FIX flags are added; plugin management is moved to the kernel
+ SDK: refresh_navband() to refresh the navigation band
+ SDK: standard file functions like fopen() are not visible by default - use qfopen() and similar functions; the standard functions can be enabled by the USE_STANDARD_FILE_FUNCTIONS preprocessor symbol
+ SDK: strarray() accepts a buffer for the answer; qstrerror() accepts a buffer for the answer; ivalue1,2,3 functions are deleted
+ SDK: take_memory_snapshot() function is added
+ SDK: term_database() function is added
+ SDK: zip_inflate/zip_deflate() functions accept 'user_data' parameter
+ SDK: added get_reg_name() to get the name of the specified register
+ SDK: bring_debugger_to_front() is added
+ SDK: dbg_trace can be used to filter trace events
+ SDK: find_text() function is exported
+ SDK: get_first_module(), get_next_module() debugger functions are added
+ SDK: new dbg_process_attach & dbg_process_detach notifications are generated in all cases ([request_]attach|detach_process() or not)
+ SDK: new function add_menu_item(), del_menu_item() (only in the gui version for the moment)
+ SDK: the asynchronous start_process() command now terminates as soon as the process is started and generates a 'dbg_process_start' notification
+ SDK: ua_emu() and ua_out() functions are not in IDA API anymore
+ SDK: user-defined form buttons can be configured in the dialog definition

post release fix: fixed problems in pro.h

Version 5.0
-----------

IDA 5.0 introduces the new graph based interface. The corresponding data
definitions and functions are in graph.hpp. These definitions heavily use
STL and virtual functions and very probably not compatible with the compilers
other than BCB v6.0.

There is a sample graph plugin 'ugraph' that demonstrates how to use the new API.

Below is the list of detailed modifications made to the API:

+ SDK: calc_bare_name() has been improved to handle __imp_ and c++ mangled names
+ SDK: guess_func_type() takes into account the number of purged bytes from the stack: if the tail parameters were not used by the function and therefore were not created by ida, we still create dummy arguments for the in the function type
+ SDK, IDC: del_segm() accepts a combination of bits as the second parameter
+ SDK: added a flag to flow_chart_t to avoid computing external blocks
+ SDK: added processor_t::gen_asm_or_lst to customize asm or lst file generation
+ SDK: added processor_t::is_insn_table_jump to determine if an instruction is really a table jump or call
+ SDK: added SDL_HIDETYPE bit for segments - to hide the segment type from the disassembly listing
+ SDK: added ui_create_tform and other callbacks to manipulate mdi child windows from plugins
+ SDK: analyze_area() function can be applied to debugger segments as well; before it was skipping them
+ SDK: areacb_t::for_all_areas() function to enumerate all areas in the specified range
+ SDK: autoIsOk() would return false for old database when called from ph.oldfile
+ SDK: callback out_src_file_lnnum to generate source file name and line number directives
+ SDK: if inf.lowoff == BADADDR, no operand will be considered as 'void' operand
+ SDK: if Namechars[] is empty, all characters are enabled in names
+ SDK: if public or weak keywords are defined as empty strings, then IDA does not display the corresponding directives at all
+ SDK: introduced new event processor_t::auto_empty_finally to handle the end of autoanalysis for efficiently
+ SDK: new function entab() to replace spaces by tabulations
+ SDK: new function qmake_full_path()
+ SDK: ph.get_autocmt notification to generate dynamic predefined comments for instructions
+ SDK: 4 new processor modules with the source code - donation of an IDA user
+ SDK: new function get_compiler_name()
+ SDK: added read_user_config_file() function

Version 5.1
-----------

We tried to keep source code level and binary code level compatibility with IDA v5.0.
We believe that old plugins and processor/loader modules will work nicely with v5.1
(there might be some bugs, of course).

However, some functions are now declared obsolote. You are still free to continue to use
them in your modules. For new modules, please define the NO_OBSOLETE_FUNCS preprocessor
symbol for the compiler. This will hide all obsolete definitions from the compiler.

If you succeed to recompile your module, it should work without conceptual changes,
unless it assumes something that does not hold anymore. For example, the nop instruction
for x86 used to be a single byte instruction with the opcode 0x90. This is not the case
anymore because Intel has introduced a multibyte nope with an operand.

We also changed the build environment a bit to support the Mac OS X.
There are also more options to fine rune the environment. See the defaults.mk
and makeopts.mk files.

The graph API has been changed as promised. Now it is (hopefully) compiler independent.

The SDK includes the source code of the Mac OS X debugger.

Have fun, and send your suggestions to us!

+ SDK: added get_bg_color callback: plugins can dynamically change the background colors
+ SDK: new function dummy_name_ea() to detect dummy names
+ SDK: added a callback to generate hints for any viewer (idaview or custom)
+ SDK: added a new group of events to detect database modifications. not all database modification generate events yet but most of them do.
+ SDK: added get_kernel_version() to get ida version
+ SDK: added save/restore pushinfo callbacks for function prolog information
+ SDK: choose2() supports column types. They are specified with CHCOL_.. bits in the widths array.
+ SDK: function to create and manipulate custom views are added
+ SDK: set_visible_func() affects all chunks of the function
+ SDK: parse_types2() allows to specify the pack alignment explicitly
+ SDK: added a callback to generate hints for any viewer (idaview or custom)
+ SDK: added a new group of events to detect database modifications. not all database modification generate events yet but most of them do.
+ SDK: added get_kernel_version() to get ida version
+ SDK: functions to create and manipulate custom views are added
+ SDK: added the qflow_chart_t class for more control on flow charts
+ SDK: added a callback for the graphs: grcode_destroyed. It is generated when the displayed graph is destroyed.
+ SDK: added a callback to preprocess flow charts before displaying them in the graph view (processor_t::preprocess_chart)
+ SDK: added a hotkey parameter to viewer_add_menu_item()
+ SDK: added add_qword() to facilitate 64-bit relocation application
+ SDK: added AFL_FIXED_SPD flag to denote instruction with fixed stack pointer delta. This flag is set at the instruction which has a fixed effect on SP (CORRECTION: it is applied to the address past the instruction)
+ SDK: added callbacks to generate custom hints
+ SDK: added decode_preceding_insn() to decode instruction which precedes the specified address in the execution flow
+ SDK: added eavec_t and intvec_t convenience typedefs
+ SDK: added events for function creation, deletion, and boundary modification
+ SDK: added FUNC_SP_READY flag to denote functions with SP analysis performed; removed ua_emu() and made automatic reanalysis of new instructions; added verify_sp notification to perform sp-analysis
+ SDK: added function_item_iterator::decode_preceding_insn() to decode instruction in the reverse execution order
+ SDK: added get_current_thread() function
+ SDK: added get_funcarg_size() to calculate the size of a function argument
+ SDK: added helper alignment template functions align_up and align_down
+ SDK: added non-standard functions like add_unique() and del() to the qvector template
+ SDK: added possibility to specify custom navigation band colorizer
+ SDK: added PR_ALIGN_INSN for some processors; this flag allows IDA to create 'alignment' instructions (e.g. nop) arbitrarily. In general it is a bad idea because such an instruction may spoil the listing by triggering the creation of other wrong instructions
+ SDK: argloc_t definition has been modified to hold more information and be easier to work with; the old definition is still available if the NO_OBSOLETE_FUNCS is not defined
+ SDK: build_funcarg_arrays() converts array arguments into pointers
+ SDK: do_unknown() and do_unknown_range() accept flags as the last parameter instead of a simple bool. This allows us to specify more options. One new option: DOUNK_DELNAMES will delete the names in the specified range
+ SDK: FUNC_PURGED_OK flag has been introduced to mark functions which certainly do not modify the stack; sp-analysis is improved to use this flag
+ SDK: gdl_graph_t has 2 more virtual functions: get_node_color and get_edge_color
+ SDK: get_data_type_size_by_flags() is optimized not to access the database if not really necessary
+ SDK: get_data_type_size_by_flags() is renamed as get_data_elsize()
+ SDK: get_off_expr and get_name_expr functions accept one more flag: GETN_NODUMMY. This flag can be used to verify that the expression can be built without really creating dummy names in the database
+ SDK: get_prev_func_addr() and get_next_func_addr() functions have been added. they return the previous/next address in the function strictly linearly
+ SDK: graph interface has been changed to use the qvector template. It is with accessible from other compilers (was only from BCB6)
+ SDK: if expand_struc() does not change anything in the structure definition because the offset is bigger than the structure size, it is not considered as an error
+ SDK: introduced a bit for user defined type information attached to program items (AFL_USERTI). Types derived by IDA will not have this bit.
+ SDK: introduced a flag for 'alignment' instructions: is_align_flow(). It is used to denote control flow created only because of an alignment instruction
+ SDK: introduced the qtype file to keep type information in the dynamic memory
+ SDK: is_type_void() now checks if the type is really 'void'; before it was reporting 'true' on partial types based on the BT_VOID symbol; the old version of this function is still available if the NO_OBSOLETE_FUNCS symbol is not defined at the compilation time
+ SDK: made graph interface compatible with compilers other than BCB6; for that, removed the bn memory of the flow_chart_t class
+ SDK: new event 'determined_main' is generated when the main() function of the application is determined
+ SDK: new event: loader_elf_machine to handle elf files of unknown machine types
+ SDK: new event: treat_hindering_item. This event allows plugins to handle the situation when an old item hinder the creation of a new item (e.g. a wrong data item prevents the creation of an instruction)
+ SDK: new function add_auto_stkpnt2() which replaces add_auto_stkpnt(). The old function should not be used anymore since it does not take into account the function boundaries
+ SDK: new function add_til2() with more flags
+ SDK: new function calc_switch_cases() to calculate switch case values and targets; jptcmn.cpp file has been modified a lot: if you used it, then take the old version from sdk5.0 or update your module
+ SDK: new function get_func_nargs(). It returns the number of arguments if the given function type
+ SDK: new function make_array_type() to create array types
+ SDK: new function qlgetz() to read zero terminated ascii strings from the input stream
+ SDK: new function set_compiler() to set the current compiler
+ SDK: new function set_purged() to specify a new value for the "purged bytes" attribute and reanalyze the relevant functions
+ SDK: new notification event: processor_t::renamed. It is generated when a new name is given to a byte
+ SDK: print_type_to_one_line() can be called with buf==NULL and bufsize==0 to find out the required buffer size
+ SDK: QASSERT() and interr() debugging helpers are added
+ SDK: QPRM() macros are useable for plugins
+ SDK: qstring template class has more member functions
+ SDK: removed HT_GRAPH and added 2 parameters to create_graph_viewer() This breaks the existing graph view plugins (sorry) but this part of IDA API has not been settled yet.
+ SDK: set_ti() and set_op_ti() always update the field list in the database
+ SDK: switch_info_t structure has been improved to handle more cases
+ SDK: tform interface was unusable because the window handles were not valid at the form creation time; added events tform_visible/tform_invisible to solve this problem
+ SDK: the kernel handles CM_CC_SPECIAL calling convention; the ph.calc_arglocs callback does not need to handle it anymore
+ SDK: the node titlebar height can be specified in create_graph_viewer()
+ SDK: ua_dodata2() replaces ua_dodata(). The new function handles correctly references to external symbols. ua_dodata() could not handle them because it had no operand offset information and could not calculate the base address of the name
+ SDK: ua_outop2() allows to specify GETN_NODUMMY ans obtain text representation of an operand without modifying the database;
+ SDK: ua_stkvar2() function accepts 'flags' parameter. Currently there is only one constant: STKVAR_VALID_SIZE. It means that the x.dtyp field contains the stack variable data type. For instructions like "lea" this flag should be cleared.

Version 5.2
-----------

The biggest change is the set of function to manipulate local type definitions.
Now local types have ordinal numbers assigned to them. This allows us to rename
them freely and get rid of name references. All references to local types are made
using ordinal numbers.


+ SDK: add_chooser_command() to add user-defined actions to chooser windows
+ SDK: added a callback for mouse click events in custom viewers; graph viewers are custom viewers too and all custom viewer functions can be used with them
+ SDK: added pc_module_t::find_reg_value callback so that plugins can find register values if necessary
+ SDK: added readsel2() to get more info about the current selection
+ SDK: added reserve(), capacity(), swap() functions to the qvector template class
+ SDK: added set_custom_viewer_range()
+ SDK: added support for negative operand values in offsets; the kernel will use a negative value if the OOF_SIGNED bit is used in outflags; PPC and M32R listings are much better
+ SDK: added wait_for_next_event() and get_debug_event() functions
+ SDK: almost all type-related functions have til_t as the first parameter; this  will allow us to introduce local type libraries and local types in the future
+ SDK: more efficient (but more memory hungry) implementation of qvector::resize()
+ SDK: new event ui_ready_to_run: occurs when the user interface is fully initialized. this event can be used to run automatic actions from plugins
+ SDK: new function compact_til() should be called before storing til file to the disk; otherwise store_til() will compact the til anyway
+ SDK: new function: parse_reg_name()
+ SDK: ph.calc_arglocs has been superseded by ph.calc_arglocs2 (optimization)
+ SDK: prototypes of type-related functions have been modified to support unlimited type strings; old functions are still available but their use is strongly discouraged
+ SDK: qeprintf() function to print on stderr has been exported
+ SDK: qsplitpath() function has been replaced by qdirname()
+ SDK: replaced value_t with idc_value_t, a class which allocated/frees memory itself and does not require manual handling
+ SDK: show_wait_box() can display dialogs with the "cancel" button. For that, pass the wait messager prefixed with "HIDECANCEL\n"
