Jean-Paul Calderone <exarkun@twistedmatrix.com> writes:
On Thu, Apr 25, 2019 at 12:52 PM Jean-Paul Calderone < exarkun@twistedmatrix.com> wrote:
Hello all,
I'd like to discuss the project policy/process for dealing with incorrect backward incompatibilities that end up released.
Considering the lack of interest in the topic, I'm going to suppose that there is no policy and the decision is made case-by-case. In the case I referenced, I'm going to approve the change in behavior as "compatible" by on the grounds that it restores the original behavior.
What you say makes sense and, as read on one of the tickets, if there *is* a need for that situation to raise an error: `AttributeError` certainly isn't the kind of error. So that should be a totally different discussion. As for a general policy: it doesn't look like this situation happens often enough for there to be a need for one (de facto, indeed that means that it is made on a case-by-case basis). Effort in general should be to prevent the situation from happening, which everyone regularly involved in Twisted does a great job at already. Basically: ACK, fully agreed, you are not talking to the void.
As a case study, we could look at https://twistedmatrix.com/trac/ticket/9410 <https://twistedmatrix.com/trac/ticket/9410#comment:17> / https://github.com/twisted/twisted/pull/1106
In Twisted 16.3.0 the behavior of Request.write was changed so that it raises an AttributeError if called after the connection has closed. Prior to that release, it silently did nothing in that case.
Now, three years later, we have a PR which proposed restoring the behavior of silently ignoring the write to a closed connection.
The original change was incompatible. The new change is incompatible. What should win out?
-- Evilham