[just ccing python-dev]
On Tue, Jun 10, 2008 at 10:17 PM, Benjamin Peterson firstname.lastname@example.org wrote:
On Tue, Jun 10, 2008 at 10:08 PM, Guido van Rossum email@example.com wrote:
On Tue, Jun 10, 2008 at 7:19 PM, Benjamin Peterson firstname.lastname@example.org wrote:
I think that we should accept Antoine's patch and begin the twilight years of sys.exc_info in favor of passing the exception instances around. This makes for more explicit and less magical code. I don't think there's any sys.exc_info case that can't be rewritten without it.
OK, assuming it works and doesn't break any unittests (or fixes the ones it expects to break), and has unittests for the new behavior, I'd say go for it.
I think the implicit chaining is assuming a little too much about the needs of the program.
That's why it's on a separate attribute. It can be handy to use when you need to debug an exception that happens in an exception handle. Sometimes it just helps to know why the handler was being invoked in the first place, other times you really want to know the original exception because that's the problem you're trying to track down. But I believe this is where Collin ran into a brick wall. I still think it could be implemented post beta 1.
Ok. I will make an issue for it in the morning.
Thanks for your pronouncements.
-- Cheers, Benjamin Peterson "There's no place like 127.0.0.1."