Re: [tclwebtest] Proposed patch to ignore socket errors

[Tilmann S. <ti...@ti...>]

* Grzegorz Adam Hankiewicz <gr...@ef...> [20030303 15:26]:
> I think that even though the user can catch any problem in
> tclwebtest's code thanks to the exception mechanism, it is less
> fearsome for users to recieve a single line human readable error
> than a complex stack trace.

If one is afraid of seeing stacktraces s/he shouldn't be programming
in tcl ;)

> Also, you might want to write specific (like null) values into
> some of tclwebtest's variables in the failure case so that other
> commands don't need to be wrapped inside catch too, or have undefined
> consequences:
> 
>  catch { do_request something_that_clearly_fails }
>  log [response text]  # <-- will this fail? what should be the output?

The behaviour after a socket error is always undefined, wether you set
all variables to null or not, that's why I still think it is good to
trow a normally uncaught error.

I rather thought catch could be used like that in individual tests if
necessary:

if { ![catch { do_request something_that_clearly_fails }] } {
   assertion_failed "Unexpectedly we got a reply from something_that_clearly_fails."
}


til


-- 
http://tsinger.com