Stack 'Used' shown in Ranger/SysMon diffs

[abalaban]

Colin:

Unless the stack is restore back to 0 after it is not used anymore, wouldn't that method indicate how much stack the task has been using *at some point* instead of the current usage?

[tonyw]

Of course - but the current stack pointer is of little interest to anyone, anyway. At the time that it is measured (by a different program), it could be anything. What you usually want to know is the maximum stack depth it has reached in the past.

[zzd10h]

Colin :
Thank you very much, I just sent you a PM.

[zzd10h]

Many thanks to you, Colin, for your StackUse source.

Stack used is now reenabled for next SysMon update.

[OldFart]

If I remember correctly, I used to fill the entire stack from the unused side upto the stackpointer with uint32's containing 0xDEADBEEF. When the program has finished it is a matter of measuring the heap of dead beef that remains. Can't tell you the exact specs as my Micro is out of order since a year and a half.

OldFart

[colinw]

If I remember correctly, I used to fill the entire stack from the unused side upto the stackpointer with uint32's containing 0xDEADBEEF. When the program has finished it is a matter of measuring the heap of dead beef that remains. Can't tell you the exact specs as my Micro is out of order since a year and a half.
OldFart

Yeah, I know this was done, I found it in a couple of old programs.
The only question I have is why, what was wrong with the default cleared stack memory value of zero. ?
A pushed return address can't ever be zero, but there's no guarantee it can't be some other arbitrary value.
It also complicates the stack testing code because it has to have a test for cookie mungers as well as for the default zero values.

[OldFart]

@colinw

Another way *MIGHT* be to find the programs Task structure and on a continuous base check the value of its stackpointer, thereby remembering the highest value reached. Ouch, all sorts of statistics could be gotten from this. Hm, I really need an Amiga soonish to sort this kind of thing out. I got that TaskView proggy somewhere on my dead Micro's HD. Might be a nice addition.

OldFart

[tonyw]

But the stack pointer in the Task struct is only updated when the task is suspended/checkpointed/waits for I/O or a signal. In between those times, it may have called a recursive function 1058 times.

The saved stack pointer is really of use only to the Exec, so that it can restore the task to a running state at some time in the future.

[OldFart]

@tonyw

Makes sense.

OldFart