Michael Gleason wrote in a message to All:

 MG> I guess I am going to show how little I still know. But
 MG> under OS/2 why would you want to / need to thunk?

 MG> OS/2 since 2.0 is 32 bit so why do you need to call 16 bit
 MG> code? Why not completely use 32 bit code?

There are several cases in which thunking is necessary.  Many people are
working with existing 16-bit OS/2 1.x programs, and may therefore have
existing 16-bit OS/2 1.x modules or DLLs.  In some cases, the DLLs may only
be available in object form, so rebuilding them or porting them to 32-bit
OS/2 is not an option.

Using certain older features of OS/2 which remain 16-bit in OS/2 2.x, such
as the Vio/Mou/Kbd API services, will require thunking, even when writing
new 32-bit programs.

There are some additional cases where people might write new 16-bit OS/2
code. For example, someone might want to use an IOPL segment, which is
considered obsolete in 32-bit OS/2, from a 32-bit program.  Also, many
people have tools which make only 16-bit OS/2 targets, such as Borland
Pascal for DOS with OS/2 kludges applied, and making 32-bit programs or
libraries would require them to replace their tools.

-- Mike


---