Hi,

If
A chunk of memory is malloc()ed.
A pointer p points to a memory location in that chunk.
The memory is then free()ed.

Can deferencing the pointer p cause a segmentation under Linux some times and
other times not.

In otherwords, can accessing memory not malloc()ed (but still in possible valid range) cause segmentation faults some times and not other times.

(Of course I know the code will be buggy, it's because I am having underterministic segmentation faults with the same program sometimes and other times not and I am trying to figure out what goes on. Note: The code is very complex.)

In case it matters, this is on a PowerPC processor.

Thanks in advance,
iTux

A segfault means that the code tried to reference memory outside one of the segments. malloc creates heap memory which is normally outside of the linker defined segments. Once the pointer is free() then
other malloc calls can reuse that memory.

The reason you don't always segfault is that the memory has been reused.
It is marked as a valid part of process space. But. There is no way to know precisely what the pointer is doing. When you segfault it simply means that you followed another code path (one that did not happen to make a malloc call that reused memory)

Short answer: yes.

iTux,

If you have a core file, you could use gdb to determine where the segmentation fault occured.

gdb <exe-name> core

then type 'where' and that will give you a stack trace..

Also try running your program through some type of runtime memory checking utility like valgrind..

Cheers...

He can also try lint