catching unexpected segmentation fault and performing cleanup

[g35rider]

Hi, sometimes due to unchecked code I get segmentation faults once in a
while and would like to be able to catch them and do some cleanup on
things. This segmentation fault could be anywhere in the code.

Could it be caught like a signal and perform cleanup then? Or would I
have to put every piece of function code in try and catch blocks?

Thanks
Ankur

 

[Ian Collins]

Hi, sometimes due to unchecked code I get segmentation faults once in a
while and would like to be able to catch them and do some cleanup on
things. This segmentation fault could be anywhere in the code.

Could it be caught like a signal and perform cleanup then? Or would I
have to put every piece of function code in try and catch blocks?

On a typical (that is all those I'm aware of!) environment, you can't
catch a segmentation/memory access fault except while running under a
debugger or emulator.

Fix your code.

 

[Mark]

On a typical (that is all those I'm aware of!) environment, you can't
catch a segmentation/memory access fault except while running under a
debugger or emulator.

It is possible to catch SIGSEGV/SIGBUS on Solaris.

Fix your code.

Agreed.

Mark

 

[g35rider]

It is possible to catch SIGSEGV/SIGBUS on Solaris.

Sorry I forgot to specify I am on Linux RH9 and FC5

Agreed.

What if it is caused by someone else's program? or library that you
have no control over, or if you are doing a client server program with
defined rules and the client breaks the rules?

 

[mlimber]

Sorry I forgot to specify I am on Linux RH9 and FC5

What if it is caused by someone else's program? or library that you
have no control over, or if you are doing a client server program with
defined rules and the client breaks the rules?

On some platforms you can do it (e.g., use signal handlers on most UNIX
platforms or __try/__except/__finally with VC++), but there is no
standard way. Ask in a newsgroup for your platform or compiler (cf. the
list at
http://www.parashift.com/c++-faq-lite/how-to-post.html#faq-5.9).

Cheers! --MM

 

[Noah Roberts]

Mark wrote:

Sorry I forgot to specify I am on Linux RH9 and FC5

What if it is caused by someone else's program? or library that you
have no control over, or if you are doing a client server program with
defined rules and the client breaks the rules?

The first should never happen.

The second you can't do anything about.

The last, your server program should never trust client code...plain
and simple. Don't access memory based on input from an external,
untrusted source.

 

[Frank Puck]

(e-mail address removed) wrote:
On some platforms you can do it (e.g., use signal handlers on most UNIX
platforms or __try/__except/__finally with VC++), but there is no
standard way. Ask in a newsgroup for your platform or compiler (cf. the
list at

in Visual C++ you can catch invalid memory access using a normal try-catch.
But you need to compile your code using the /EHa option.
There is also a callback which you can set to throw your own type of
exception.
This callback must be set using _set_se_translator.
Otherwise catch(...) must be used.