Re: GNAT's stack checking in Ubuntu 9.04 (and Shootout regex-dna)

[Georg Bauhaus <rm.dash-bauhaus@futureapps.de>]

Vadim Godunko schrieb:
> On Aug 4, 1:17 pm, Georg Bauhaus <rm.tsoh.plus-
> bug.bauh...@maps.futureapps.de> wrote:
>> 8192 is the default output of ulimit -s on Ubuntu.
>>
>> I had set ulimit -s $((8192 * 64)) on both Ubuntu 9.04
>> and Debian 5 and both exhibit the same symptoms.  Why?
>>
> This size is too small,

What's the reason?

> you can setup unlimited size by
> 
> ulimit -s unlimited

The machine happens to have 500 Mo of RAM; I had
thought that would be sufficient for 2-3 times
50 Mo of input when processing is assumed to be
mostly linear (at worst doubling each string).
The program does fine in ulimit -s 8192 when the
Dna_Lines array is allocated on the heap, so the
overall amount of memory is not the problem.

It hits the limit exaclty when the array of 625_000
default initialized Vstrings is allocated on the stack.
No operations performed at that point.


> Unbounded_String uses constant size of stack for object itself and
> uses dynamic size on heap for actual value.

Yes. How much does a default initialized Unbounded_String
use when it is an array component? (I'll debug this when
there is more time; just in case you happen to know,
would be most helpful.)


> But note, concatenation
> operations of strings and arrays can use stack for intermediate
> result, thus can be the source of problem.

This was one of the fixes for the K-Nucleotide test.
(Do the lower -> upper character mapping in A.S.Unbounded,
then call To_String; not: call To_String and feed that
to To_Upper.)