Subj : RE: Re: Possible FP87/EMU Problem?
To   : borland.public.cpp.borlandcpp
From : kc <kc_22178@yahoo.com>
Date : Wed Jan 14 2004 05:35 am

I am working in a C & assembly enviromment. In this case can you let me know how can I check the size of stack so that I would not run into stack overflow situation.

kc


>"Wayne A. King" <waking@idirect.com> wrote in message
>news:3ad23dde.4789130@newsgroups.borland.com...
>>
>> Are you building with the option to check for stack overflow enabled?
>> Have you tried increasing the stack size via the _stklen variable?
>> (Assuming it’s a DOS application.)
>>
>
>We have the option checked for stack overflow. We never receive a stack
>overflow error. I am going to put in some code to examine the stack area to
>see how much we’re typically using. We have not increased the stack size
>with the _stklen variable yet.
>
>That raises a question. How can you change the stack size at run time with
>the _stklen variable, when it would seem that the compiler really needs to
>know this during the link (i.e. it reserves a certain amount of area for the
>stack).
>
>Yes, it is a DOS application. In fact it’s a real-time application that
>communicates with a custom serial I/O board. It’s impossible to run with any
>sort of debugger, given the nature of the hardware and software. And due to
>the way the problem manifests itself, it’s been difficult to insert any
>helpful code, since the simple insertion of a line of code often makes the
>problem disappear.
>
>Get this - if I insert an empty 75 character array at the very end of our
>software (it gets inserted by the linker at the end of bss, just prior to
>the stack) the software appears to work fine. If I make it a 1 character
>array, the software breaks. But if I make it a 1000 (or 2000, or 4096, or
>almost anything else) character array, the software breaks too. Fooling the
>linker into moving the stack around changes the problem, but it’s not
>necessarily better when you move the stack a longer distance from our
>application.
>
>We are certainly concerned about stack overflow, although it didn’t seem
>likely since it seems it would have been caught due to the run-time stack
>checking being enabled. I’ll be looking into that more tomorrow. We’re
>running out of ideas, however.
>
>Thanks...
>
>Andy Johnson
>
>
>
>
>

Message sent from http://www.aewnet.com

.