Re: eclipse 5.5 gives error sometimes?

On Wed, 2002-11-27 at 12:04, Joachim Schimpf wrote:
> Sebastian Sardina wrote:
> > 
> > I installed eclipse 5.5 in two red hat 8 machines. in one, everything
> > works fine. in the other one I get:
> > 
> > [ssardina@cogrobo ssardina]$ eclipse
> > 
> > *** ECLiPSe fatal error: Cannot reserve stack space in alloc_stack_pairs()
> 
> I don't have redhat 8 and cannot reproduce this on any machine
> we have available.
> 
> A freshly started eclipse should consume less than 5MB of memory
> (by the way, real memory is irrelevant here, what counts is
> virtual memory, i.e. swap space).
> 

yes you are right, it's total memory that matters. 

> When you run top(1) or free(1) you should see an increase in
> memory usage of max 5MB for every eclipse you start. If you
> see anything different, let us know.
> 
> 
> > strange, I did an strace and here is where the error seems to be:
> > 
> > mmap2(NULL, 268435456, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = -1
> > ENOMEM (Cannot allocate memory)
> 
> Eclipse does this call in order to reserve address space for the
> stacks (default 2*128MB). This call should not actually consume any
> memory or reserve swap space, at least it doesn't in the Solaris and
> Linux versions we use for testing. I hope this has not changed
> in recent Linux kernels... It might also be a bug in Redhat 8.
>
>
> > It seems my machine (64Mb) does not have sufficient memory to run eclipse
> > as it is.
> > If I use -g 50M to set the global stack to 50Mb, then it works!
> > 
> > shouldn't eclipse recognize how much memory the machine has available?
> 
> As I said above, the only limit should be swap space, not real
> memory. Moreover, the memory for the global stack does not get
> allocated all at once, but incrementally as the stack grows.
> If that should be broken in Redhat 8, we need to look into it.


ok I will investigate this. My machine has 64MB of RAM and 400MB of
swap. I think that at the time I ran ECLIPSE I had 192MB of virtual
memory free so I thought that 2*128 > 192 was the cause. But if, as you
said and is natural, ECLIPSE allocates memory incrementantally it should
have run. I will investigate this further.

In any case, I could run eclipse using -G 70M, anything higher breaks.

Sebastian



> -- 
>  Joachim Schimpf              /             phone: +44 20 7594 8187
>  IC-Parc, Imperial College   /            mailto:J.Schimpf@ic.ac.uk
>  London SW7 2AZ, UK         /    http://www.icparc.ic.ac.uk/eclipse
>