MadSage wrote:

Currently I don't do anything in the catch block. If I get an exception

one of my threads it stops running, allowing other threads to continue.

I would like to restart the thread. Would it be ok to put the try catch
blocks in a while block like the code below? Will this allow the thread

continue, providing the original bug which caused the exception doesn't
reoccur? Obviously I should log the exception too so I can look into

but I would like my server to continue running if at all possible.

You should be sure to diferentiate a bug and an exception.
If you got an exception because something that you might have expected
to fail in extreme circunstances (memory allocation, io error), then it
is ok to retry it after fixing the problem, if it can be done.
If you got a bug there, then it is better to have it fail loudly so you
can look at it. In my view, this will prevent a lot of future headache.

Marco Costa

MadSage wrote:

I currently have a multi-threaded server application with worker threads

a core thread. For all of these threads I have something like the

code:

uint32 __stdcall CCore::EntryPoint(void *pV)
{
CCore *pCore = (CCore *)pV;

try
{
while(true)
{
pCore->Update();
Sleep(20);
}
}
catch(...)
{
}

return 0;
}

Currently I don't do anything in the catch block. If I get an exception

one of my threads it stops running, allowing other threads to continue.

I would like to restart the thread. Would it be ok to put the try catch
blocks in a while block like the code below? Will this allow the thread

continue, providing the original bug which caused the exception doesn't
reoccur?

There's no inherent problem with having the try-catch in a loop, and
the multithreaded-ness of the program seems to be irrelevant as far as
your question is concerned. (In general, you shouldn't allow
execeptions to propogate outside your code to a caller you don't
control, such as when you utilize OS callbacks like it appears you are
doing here, and, likewise you should make sure each thread catches all
of its exceptions since otherwise they will propogate to OS-specific
threading code.) The real question is, "Is it safe to re-execute
CCore::Update() after an *unknown* exception is caught?"

Obviously I should log the exception too so I can look into it,
but I would like my server to continue running if at all possible.

[snip]

Right, but you'll have to catch a specific exception class (perhaps
using a reference to a base class like std::exception) to get any
information out of it. <OT>Also consider that Microsoft uses
"structured exception handling" to make certain errors (e.g., access
violations, stack overflows, etc.) look like exceptions. You can handle
these with a catch(...) or with their non-standard
__try/__except/__finally keywords. Check out your documentation for
more.</OT>

Cheers! --M