Andrew Dunstan <andrew@dunslane.net> wrote:
> My dim recollection is that Tom and I and maybe some others did
> tests on a bunch of platforms at the time we introduced the
> protocol to make sure it did work this way, since it's crucial to
> making sure we don't get interleaved log lines.
Testing is good; I like testing. But I've seen people code to
implementation details in such a way that things worked fine until
the next release of a product, when the implementation changed. I
was surprised to see Tom, who is normally such a stickler for doing
such things correctly, apparently going the other way this time; but
it turns out that he had noted a guarantee in the API that I'd
missed. Mystery solved.
Perhaps something in the comments would help people avoid making the
same mistake I did.
-Kevin