Re: [RFC] Fission patch 1/2

[Tom Tromey <tromey@redhat.com>]

>>>>> "Doug" == Doug Evans <dje@google.com> writes:

Doug> I wonder if to some, but not complete, extent (*1) cleanups are more
Doug> fragile than necessary because the API is more fragile than necessary.
Doug> My intent was the opposite, but ok, such is life.

In general I think making them more dynamic is troublesome.  It makes
the logic of the function harder to read.  I'd rather cleanups be even
more static, like C++ destructors.  Then in the cases where dynamic
behavior is needed, it would be done locally via particular local
objects, rather than manipulations of the cleanups themselves.

Tom> It is better to keep a separate flag.

Doug> That sounds pretty odd (and error prone).

Yeah.

Doug> Are there *useful* situations in which make_cleanup can return
Doug> NULL?  Is it only the first one?

TRY_CATCH resets the cleanup chain.  So, I think the first call to
make_cleanup inside any TRY_CATCH will return NULL.

Doug> It feels like it would be cleaner if that were never true, and
Doug> thus the users needn't have a separate flag, and thus can be
Doug> simpler (and thus the intuitive choice isn't the wrong thing to
Doug> do).

I agree this would be an improvement.  Though I'd still prefer that most
code not play games with cleanups at all and just treat them as
block-scoped as much as possible.

Tom