Re: RFC [Discussion]: Improve unserialize() error handling

Hi On 9/28/22 13:41, Christoph M. Becker wrote:

Predicting people's second-place choice is risky business. This assumption seems logical on its face, but I'm sure there are people that will buck your expectations.

The reasoning is that unless “E_WARNING in 8.x without future decision” receives more than 50%, more than 50% prefer an Exception no later than 9.0. Unless “UnserializationFailedException in 8.x” receives more than 50%, more than 50% prefer no Exception in 8.x.

If you want to go that route, I'd go all the way to an RCV vote and be done with it. Or else just make an executive decision as the RFC author and let the chips fall where they may.

I'm generally not too happy with secondary votes. Sometimes you only support the primary vote for certain secondary options; to "be sure" that another secondary option won't "win", you'd need to vote "no" on the primary choice. I'd prefer a single vote with pre-selected details. I don't have any particular preference in this case, though.

Would the increase of E_NOTICE to E_WARNING for the two cases that currently emit E_NOTICE be something that even requires a vote or is this something that can simply be decided by merging a PR? [1] In that case the second and third vote could be simplified to "Convert E_* to Exception" with the regular 2/3 majority required. [1] I would hope that unifying E_NOTICE/E_WARNING to E_WARNING everywhere is uncontroversial. Best regards Tim Düsterhus