On Friday, 1 May 2015 at 07:06:53 UTC, Andrei Alexandrescu wrote:
On 4/30/15 6:12 PM, Daniel Murphy wrote:
"Andrei Alexandrescu" wrote in message
news:[email protected]...
Nothing negates that. It's a judgment call. Please let's stop
here.
Thanks. -- Andrei
Please stop trying to undo this improvement. Just fix your
code and
move on.
Thanks.
After we have discussed this matter extensively Walter and I
decided to move forward and undo this language change.
It's a judgment call. The advantages and disadvantages of this
have been discussed extensively. We are accountable for
decisions like this, and we must make the decisions we believe
are right. That said, making a decision when there are strong
feelings and opinions on both sides is always difficult.
We would really like to find consensus, but failing that,
request your indulgence.
Please review:
https://github.com/D-Programming-Language/dmd/pull/4623
Thanks,
Andrei
You say you want to find a consensus, but neither you nor walter
provide a good sum up of your discussion.
That makes it hard to express any good rebuttal or even be
understanding of you position.
People here have presented situations where they faced bugs due
to the current behavior, and even Walter got it wrong in the doc
in the past (which makes any point saying that the current
behavior is desirable moot).
Right now, the main reason expressed on your side is the breakage
in std.allocator, which is known to be fairly different than your
usual D code. It lets most people here with the feeling that
breaking our code is no big deal (I have breakage in SDC with
pretty much every release, usually for the better, sometime
completely for nothing like when struct construction were not
lvalues anymore) but breaking yours is.
It has been presented in this thread how the current behavior
lead to confusion (including, as already mentioned, from Walter
himself) and how unit tests is likely to NOT catch it, because it
create this kind of situation where things mostly works, but
really do not.
The only person here that has presented convincing evidence
against the change is Vladimir, and that's be really interesting
to dig into his report to know if all the reported uses are
correct or if the change discovered some bug in his code.
Possible less painful transition path also have been proposed,
but none has been discussed.
I sure have made the mistake myself several time and have to make
people fix their PR because of incorrect use of slice to bool
conversion on a regular basis, so I'd expect that this damn thing
to be a positive change. It seems that the majority of people
agree. Granted, it is not because most people think something
that it makes it right, but that certainly is a good
justification to at least provide detailed explanation of why the
majority is wrong if it is the case.
