WP 2.9: three bugs, and how to fix them…

Thanks for the link. I’m already passing that on when appropriate. FWIW, I’ve also come across 1 or 2 instances where similar symptoms (missing scheduled posts) seem to have been caused by an incomplete or corrupt upgrade that was fixed by a re-upload of wp-admin and wp-includes.

Another thought: since both of the issues addressed by the patch seem to relate to server issues, should the Hosting Requirements page in the Codex be updated to include references to the iconv and multi-byte string functions and the minimum cUrl library required?

Re the codex page, it certainly won’t help to add a notice.

I’m not sure it should be mentioned as an outright server requirement, though — even if it de facto has become such a thing. The committers are not very far from patching the cron issue, and I’m pretty certain thatthe SimplePie-related problems will get fixed in 2.9.1…

Because I lose track of these things so easily… saved a link at https://codex.www.ads-software.com/User:Miqrogroove

??

Why is something released it’s users’ as stable and forced upon them to uprgade, when it contains 3 not-so-small bugs? These things should be caught during development, and if not then definitely during testing..

@xdesi: my own take on this is:

https://westi.wordpress.com/2009/12/19/what-should-go-into-a-wordpress-maintenance-release/#comment-13632

Westi definitely doesn’t agree, though. ??

Westi opened that thread in a completely biased introduction, and has so far disagreed with every reply to it. I think that speaks for itself.

dd32 has a much more approachable attitude https://dd32.id.au/2009/12/21/a-call-to-arms/

I agree with xdesi. The issue isn’t that buggy stuff gets released, it’s that buggy stuff gets released as stable. When a stable release is unstable, that makes WP look bad. Some people like to mess around with unstable stuff and try it out, lots of us don’t need that. We don’t have time to hack code and maintain content on the website at the same time. Last version was fine, there was no reason to rush 2.9.

#11168 went through over a month of testing and nobody triggered the error.

On #11219 the maintainer of the third-party library in question disagreed with the proposed fixes. The concern was that fixing a bug that affects a small number of people could break something for a broader audience.

#11505 was a good, old-fashioned screw-up. My fault.

Note that two of the bugs are very dependent on server configuration, with one of them being triggered by a bug in certain versions of PHP. These bugs can go through an eternity a beta testing without anyone ever noticing them because the beta testing audience doesn’t include anyone with that particular setup. These bugs often don’t surface until the final release, which sees an endless variety of server configurations.

Regardless, one bug is already fixed in the 2.9 branch, another will be shortly, and I’m trying to get feedback from the third-party library maintainer.

#11168 went through over a month of testing and nobody triggered the error.

I don’t mean to rub it in or anything, but when the HTTP API is broken, it becomes a lot easier to find unhandled errors in SimplePie ^_^

To be fair, I think the default timeout for that class is 10 seconds, so probably had nothing to do with it.

…

@ryan: “The concern was that fixing a bug that affects a small number of people could break something for a broader audience.”

If that really was his concern upon viewing the two patches, then I am simply speechless.

The first patch I added merely amounts to saying: “We’re trying to convert UTF-8 into UTF-8, so let’s just assume that it works.”

It got dismissed pending further investigation and potential security implications, which is fair enough. But then the second patch goes: “We can’t validate UTF-8 from lack of server functions, and we’re trying to convert UTF-8 into UTF-8, so let’s just assume that it works.”

Quite honestly, I think that the maintainer of SimplePie didn’t head straight to the lines in question to wonder what I was adding.

I’d also like to stress what I wrote on Westi’s blog post. It’s totally discouraging to see this kind of thing ignored.

I’m actually quite happy that I told my customers to not upgrade until 2.9.1. I felt a knot in the stomach when I did, since I was liking what I was seeing in WP 2.9, but with over 1k pending tickets in trac, I knew something like this would happen. I’d currently be full time on upgrade support if I had not done so.