[cc-ing Laszlo-Dev, in case other people want to chime in.]
On 2009-09-14, at 19:56, James Robey wrote:
I want to define bomb silently again, to be sure you understand;
bomb silently means that the runtime halts the initialization stage
at the time of error, often leaving applications completeIy blank
and unresponsive, depending on how deep in the API the problem
occurs. There is no debugger warning. I've notice a large group of
situations where the runtime will halt in this fashion, and because
i'd like to help i'd like to ask the following question:
Are all of these bugs to be reported?
I think that this bug:
http://jira.openlaszlo.org/jira/browse/LPP-8151
probably is sufficient for the "bomb silently", problem for now.
I went back and looked at the state of debugger support for catching
errors and see that I only enabled it in the case where backtracing is
on, but we don't support backtraces in swf9 yet, so that's basically a
no-op. In DHTML, you have a similar situation, there are errors that
will halt Javascript in the browser that would not cause an error in
swf8. In DHTML, you have to resort to the browser debugger (Firebug,
Webkit's built-in, IE's add-on), just as you have to resort to the
Flex debugger in swf9 to track these down. In DHTML, I tried to help
a little bit by enabling the catching of errors automatically when you
turn on backtracing (since backtracing needs to establish try/finally
blocks anyways), but... You'll see that I am prevented from having
the debugger always catch errors and report them by:
http://jira.openlaszlo.org/jira/browse/LPP-8222
so I am trying to come up with a compromise. It will probably be
something along the lines of Java, where if you mean for a method to
be able to throw a runtime error, you have to declare it so.
