[A92] Re: TI92+ HW2 AMS 2.05 ASM size limit
[Prev][Next][Index][Thread]
[A92] Re: TI92+ HW2 AMS 2.05 ASM size limit
On Wed, Aug 08, 2001 at 05:33:29PM -0700, Patrick Davidson wrote:
>
>
>
> On Wed, 8 Aug 2001, Rudolf Polzer wrote:
>
> >
> > On Tue, Aug 07, 2001 at 04:11:44PM -0700, Patrick Davidson wrote:
> > >
> > >
> > > On Tue, 7 Aug 2001, Rudolf Polzer wrote:
> > >
> > > > Isn't the ASM size limit 24k on HW2, AMS 2.05?
> > >
> > > Yes, it is.
> > >
> > > > I have installed the HW2Patch, but my TIGCC program (22k) still
> > > > crashes. Interestingly, it crashes before my code is executed (the
> > > > first line to be executed is a clrscr()!) and I get an "Address
> > > > Error". If I run my program (compiled in Doors mode) in UniversalOS,
> > > > it works.
> > >
> > > The problem isn't likely due to the size limit. If the program were over
> > > the size limit with no kernel installed, you'd get the "ASAP or Exec
> > > String Too Long" error message. On HW2, there is a hardware protection to
> > > back up the size limit by prohibiting code being executed in certain
> > > address ranges, which can cause crashes if you try to run code outside the
> > > range without circumventing or deactivating the hardware protection (which
> > > HW2Patch does for you anyway). In any case, unless your program tries to
> > > run code somewhere in RAM outside its own boundaries, this won't happen,
> > > and if it did the result would not be "Address Error" .
> >
> > My code cannot be the fault here - the first thing I do is a
> > clrscr(), but the screen is not cleared before the call.
> > Also, if I leave out the big (11k) .o file and replace
> > it with a small one providing the same exported function
> > (with empty body), it works. But I call this function
> > _after_ getline() (TIGCC example code).
> >
> > Maybe a bug in the linker? In the compiler? Is there a
> > working 9XZ disassembler?
> >
> > I use no linker flags and '-Os -fno-omit-frame-pointer' as compiler
> > flags.
>
> If I had to guess, I would say this is most likely a bug in the linker,
> though I'm not certain of it. This is especially likely if the .o file is
> in the Amiga object file format (as generated by A68K) because the linker
> doesn't do too god a job of combining such files with coff format files
> (which are made by GCC and the GNU assembler).
All are made by gcc. But AFAIK it calls A68k.
I have another guess: maybe the startup code assumes that the distance
to main is lower than x or something like that (# global variables,
functions etc.)?
> One 9xz disassembler is included with the TIGCC Tools Suite; it is listed
> as an early beta but it seems to work well. Also you can use the debugger
> of VTI which also allows you to run through the execution step by step.
I'll try VTI's debugger - I do not find my main function in the
disassembly. Is there something like a debugger trap (like INT
3 in DOS)?
Follow-Ups:
References: