Catch bad_alloc?

[kk_oop]

I'm looking for opinions for or against catching std::bad_alloc when I
use the "new" operator. I searched through this newsgroup and the
Internet, and I don't see clear opinions either way. Note that I am
using gc++ version 3.2 on a unix platform, so I'm not concerned with
any points unique to VC++.

Thanks!

Ken

 

[Ben Pope]

I'm looking for opinions for or against catching std::bad_alloc when I
use the "new" operator. I searched through this newsgroup and the
Internet, and I don't see clear opinions either way. Note that I am
using gc++ version 3.2 on a unix platform, so I'm not concerned with
any points unique to VC++.

Depends if you want to handle memory errors. Generally it's a good
idea, but often you can't do very much but close the program. It's
always nice to tell the user that they've run out of memory before
crashing hard .

It's possible you can dump some other data, to free up memory and
continue. It's your call.

Ben Pope

 

[Markus Moll]

Hi

It's possible you can dump some other data, to free up memory and
continue. It's your call.

This is better done in a new_handler, imho.

Markus

 

[Bo Persson]

Hi



This is better done in a new_handler, imho.

That's good at the global level. In a layered application, it might be
better to have a separate catch statement at some of the intermediate
levels.


Bo Persson

 

[Roland Pibinger]

This is better done in a new_handler, imho.

Agreed. OOM (Out of Memory) conditions are usually caused by memory
leaks. You cannot catch or handle memory leaks so it's best to
gracefully shut down the program in the new_handler. See also:

http://www.gotw.ca/publications/mill16.htm
http://www.baus.net/memory-management