Цитата из комментария в коде, находящемся по первой (!) ссылке в Гугле (запрос "FT_Thunk QT_Thunk":
/* The set of FT_... thunk routines is used instead of QT_Thunk,
* if structures have to be converted from 32-bit to 16-bit
* (change of member alignment, conversion of members).
*/
---
update: И еще оттуда же:
******************************************************************** * FT_Thunk (KERNEL32.234)
*
* This routine performs the actual call to 16-bit code,
* similar to QT_Thunk. The differences are:
* - The call target is taken from the buffer created by FT_Prolog
* - Those arguments requested by the thunk code (by setting the
* corresponding bit in the bitmap at EBP-20) are converted
* from 32-bit pointers to segmented pointers (those pointers
* are guaranteed to point to structures copied to the stack
* by the thunk code, so we always use the 16-bit stack selector
* for those addresses).
*
* The bit #i of EBP-20 corresponds here to the DWORD starting at
* ESP+4 + 2*i.
*
* FIXME: It is unclear what happens if there are more than 32 WORDs
* of arguments, so that the single DWORD bitmap is no longer
* sufficient ...
*/
Цитата из комментария в коде, находящемся по первой (!) ссылке в Гугле (запрос "FT_Thunk QT_Thunk":
/* The set of FT_... thunk routines is used instead of QT_Thunk,
* if structures have to be converted from 32-bit to 16-bit
* (change of member alignment, conversion of members).
*/
---
update: И еще оттуда же:
******************************************************************** * FT_Thunk (KERNEL32.234)
*
* This routine performs the actual call to 16-bit code,
* similar to QT_Thunk. The differences are:
* - The call target is taken from the buffer created by FT_Prolog
* - Those arguments requested by the thunk code (by setting the
* corresponding bit in the bitmap at EBP-20) are converted
* from 32-bit pointers to segmented pointers (those pointers
* are guaranteed to point to structures copied to the stack
* by the thunk code, so we always use the 16-bit stack selector
* for those addresses).
*
* The bit #i of EBP-20 corresponds here to the DWORD starting at
* ESP+4 + 2*i.
*
* FIXME: It is unclear what happens if there are more than 32 WORDs
* of arguments, so that the single DWORD bitmap is no longer
* sufficient ...
*/
---
никто не знает?14.03.04 13:00 Автор: Kerk Статус: Незарегистрированный пользователь