From: Andreas K. <and...@ac...> - 2010-03-31 17:15:43
|
Donald G Porter wrote: >>> So, is this a bug in the implementation, or is this the behavior >>> desired by the proposers of TIP 348, which seems to be leaving out >>> the benefit at possibly the most useful point? > > Alexandre Ferrieux wrote: >> It is a limitation due to the usual complexity-power tradeoff. > > Yes, I understand all that. > > What I do not know is whether or not, with this limitation, TIP 348 > really answers the requirements of folks who have been asking for it. > I know you are satisfied, Alex. Anyone else? > > Anyone else tried applying the patch, testing your problem cases and > seeing whether it solves the problem? Well, the TDK Debugger in its current form is not affected by this either way. It instruments the code, effectively wrapping each and every command into a [catch], giving it full and total control over the stack unwinding, including the ability to not unwind and ignore the error. The errorInfo information does not come into play there. There I (can) see the benefit of -errorStack is in the more interactive context, i.e. tclsh, wish console, tkcon. The stack is (already) unwound and we are in post-mortem mode, trying to determine what went wrong after the fact. For this situation having this last half-level would indeed be beneficial. OTOH, if I understand Alex description correctly, due to how this works internally (ref 'inner context'), even if we manage to put this half-level into -errorstack it would be no different from the information we have in -errorInfo already, i.e. the command with unsubsted arguments. As such I could live with the current implementation, simply looking at both errorstack and errorInfo during debugging. My 2 cents... -- Sincerely, Andreas Kupries <an...@ac...> Developer @ <http://www.activestate.com/> |