Re: How to deal with bugs in vendored libraries?

One good way to maintain some quality standard is to enforce it thru CI :) This also applies to https://github.com/php/php-src/pull/10549/files and https://github.com/php/php-src/pull/10531/files for example, such problems should fail the CI and enforce the contributors to fix the code before the changes are merged. With kind regards / Mit freundlichen Grüßen / S přátelským pozdravem, Michael Voříšek On 9 Feb 2023 14:14, Max Kellermann wrote:

On 2023/02/09 13:37, Rowan Tommins <rowan.collins@gmail.com> wrote:

Firstly, let's try to keep this discussion civil, and assume good faith on both sides. Parts of your e-mail read like accusations of bad behaviour, rather than genuinely trying to understand what happened, and how we can collectively avoid it happening in future.

While I do have an opinion on whether I consider Derick Rethan's behavior bad (yes, I do), that's not the point here. I don't know how to proceed after my PR thread was locked - that's an unequivocal sign of refusal to discuss the issue. The issue still exists, and I'm here for your advice on how to resolve this. I'm desperate.

Secondly, note that Derick Rethans is the maintainer of both timelib and the ext/date extension in php-src. So while we can discuss the hypothetical question of how to handle a disagreement between php-src and upstream library maintainers, it wouldn't apply in this case anyway, because it would require Derick to disagree with himself.

That depends. Did the PHP project decide to go C99 starting with version 8? What does that mean for maintainers - can they decide to make code changes that are not compliant with that decision? That is a honest question. I don't know how PHP works. That's why I asked whether "secret" reverts without discussion are considered good behavior. Maybe you believe maintainers should do that - but that would be surprising for me.

Thirdly, it's not clear to me which of the following statements is true of this change, and it might help the conversation to clarify more precisely: a) The code you removed *violates* the C99 spec?

This. The code in question declares typedefs that are reserved words in the C99 spec section 7.26.8; not just reserved, they conflict with actual typedefs from <stdint.h>.

b) The code you removed is *guaranteed to be pointless* under the C99 spec (but does not violate it)?

No. It is not pointless. Those declarations occupy reserved words, and that is not allowed.

c) The code you removed is *pointless in this particular case* because of a combination of the C99 spec and other factors (but might be reasonable in other circumstances)?

I don't understand this one, but it doesn't sound like it applies. Max