Very Computer

  In almost every book about C++ I read, they
  say that, if 'new' fails, it returnes a pointer to
  NULL or they throw an exception (std:.bad_alloc).

  Unfortunately, I compiled my application on a
  outdated woody snapshot with gcc 2.95.3 and
  I was not able to prevent my code from crashing.

  When the system memory was exhausted, the
  programm was aborted and a message 'out of
  memory' appeared on the console.

  I encapsulated the dangerous code in a 'try'
  block and I catched 'std::bad_alloc' as well
  as every other exception (catch(...)) but this
  didn't help.

  Testing the pointer if it is NULL didn't help
  either.

  Is there a third way of what could happen
  if 'new' fails? I only know the NULL- and the
  exception thing.

  On my machine at home (2.95.4, woody)
  an exception is thrown and can be caugth.
  But on that other woody machine, I couldn't
  handle this situation.

  Martin

--

http://www.martinhenne.de
http://www.aachenprogrammingclub.de
http://apcstudio.sf.net

Quote:>   Is there a third way of what could happen if 'new' fails? I only know
>   the NULL- and the exception thing.

I've found that STLport (http://www.stlport.org) helps GCC's C++
conformance considerably.  While it's not germane to the exception problem
you're facing, if you do any sort of serious C++ work I'd strongly suggest
ditching GCC's STL as soon as possible.

STLport is free software, available under an MIT/X11-style license.

You will probably see something like this on the console:

  kernel: Out of memory, killing `yourprocessname' (pid 1234)

It should also appear in `dmesg'.

--

"I woke up this morning and realized what the game needed: pirates,
pimps, and gay furries."  - Rich "Lowtax" Kyanka

--
Kasper Dupont