Eugene Gershnik wrote:
> On Apr 30, 9:43 am, "Martin T." <0xCDCDC...@[EMAIL PROTECTED]
> wrote:
>> Hi all,
>>
>> I'm currently trying to come up with a decent exception base class for
>> some modules of ours and was thinking about adding exception-chaining
>> (yes, like in Java :-)
>> The following points where im****tant for me:
>> * Must inherit from std::exception
>> * Should allow for a std::exception to be the start of the exception
chain.
>> * throw by value / catch by const reference
>
> I have been using a similar solution with similar goals for quite some
> time now and so far pretty happy about.
It's good to hear that others are using something like it. Have been
wondering why I haven't found more information about this ...
> The problem with your implementation IMO is that you burden every
> class derived from chained_exception with often unused members that
> contain information about std::exception based 'cause'. What I did to
> avoid this was to have a dedicated type, let's call it,
> std_exception_wrapper that did the job of extracting type_info, what()
> from std::exception and exposed the same chained_exception interface
> with clone() etc. The chained_exception itself ends up holding a
> single pointer to possible cause.
>
> The usage would be
>
> catch(std::exception & e)
> {
> throw chained_exception(...., std_exception_wrapper(e));
> }
>
As I would like for adding more "debugging" information to the exception
base class (file, line, function, info about current callstack) the
exception-wrapper or an exception-info class will probably come in
handy, yes.
> Another thing I've done differently is to have a generic 'printer'
> facility that allows me to do
>
> std::cout << exception_printer(ex, <some params that might make
> sense>);
>
> This is more flexible than having a 'print' member.
>
Sound very neat. I find that to much genericity tends to annoy my
co-workers for some reason, so I might stick with a few focused
functions/methods :-)
> If you don't strive for universal ****tability and only intend to
> sup****t major modern platforms you also might want to store backtrace
> information in the chained_exception. Makes debugging remote
> installation problems even easier. ;-)
>
Well, on this side we're a Windows only shop so at least I do not have
to care too much about ****tability.
Thanks for sharing your experience!
cheers,
Martin
--
[ See http://www.gotw.ca/resources/clcm.htm
for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]
|
