On Thursday, 30 August 2018 at 23:03:57 UTC, Walter Bright wrote:
On 8/25/2018 4:49 PM, Nicholas Wilson wrote:
Run semantic3 on the constructor independent of the
requirement to destruct already constructed objects. If the
constructors is nothrow then there is no need to have the
destructors run or the eh code at all, because no Exceptions
can be thrown (an Error may be thrown but that will kill the
program). This is how I intend to fix it after I refactor
semantic3.
A function can be made nothrow by:
try {
....
} catch (Exception e) {
... handle it locally ...
}
Then I should have said: no exceptions can propagate, which is
the real problem.
Also, your proposal is ignoring the destructors, which is
literally what the compiler does now.
It was implicit in that the throwing case would call the
destructors in the event of an exception (otherwise the bug ain't
fixed). This formulation is to reduce the amount of breakage,
which was the problem last time.
Yes this will break (as in code breakage) @safe ctors calling
@system dtors but, such is life. The ctor probably shouldn't be
throwing in the first place. I'll probably add -vthrowingctor and
-vthrowingdtor as well since this will be a perf hit in the case
of a throwing ctor.
Sorry for any confusion.
