newlib malloc problems

[Jason Wilkins <fenix at io dot com>]

I somehow sent an incomplete version of this message.  Please disregard.

I am using newlib malloc to manage a small heap of just 256k.  One of my 
users has sent me a pathological test which causes malloc to fail.

Using C++, the test program allocates 1000 arrays of 42 integers, then 
deletes them all, then attempts to allocate a 64 kilobyte array.  The 
allocation of the large array fails because malloc attempts to use sbrk to 
allocate new memory for the array, but there is not 64k left at the end of 
memory.  It should not attempt to use sbrk however, because all the memory 
that was allocated before should be freed at that point and it should 
have allocated from the free space.

>From my reading of the source code of 'free', when all the small arrays 
are deallocated, they should have been combined into a progressively 
bigger free chunk until they were just one giant free chunk.  They are 
freed in the same order that they were allocated.  

Then, malloc should have seen that there was a sufficiently big enough 
chunk to allocate from and put the big array into it.  Instead, it seems 
to be failing to find the free space and is trying to sbrk some more.

It may be possible that this is a C++ problem, so I am going to rewrite
this test in C to see if there is something funny going on with the C++
new[]/delete[] operators.



------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com