Thks Grandpa taking care of the babies.

You cannot update the extra_query after an update has been discovered

Is addressed here #32862

OPcache problem

For Joomla 4 is here #32915

For Joomla 3 is here #32918

For joomla-fframework/filesystem here joomla-framework/filesystem#37

@PhilETaylor I hope you know that Sharky doesn't have any right to speak for the JBS or for maintainers. He seems to perform his private war here. So please don't let him influence your motivation for your PR's.

I will use have used official channels to escalate this to OSM. Who I fully expect to do nothing. As they have done for every single CoC issue I have raised. @jeckodevelopment

Hi @PhilETaylor calling @marcodings here as the relevant DC.

@jeckodevelopment @marcodings More information on what the problem was can be found in this PR: #32915 . You have to expand the comments which are meanwhile hidden on GitHub by default due to the long history. You can also ask me on Glip for details.

due to the long history.

Which is me actually trying to get this right and get it all merged as the opcache issues are valid issues already addressed by all other major cms. And thank you for @richard67 for his timely feedback and time taken to actually understand and test the issues involved.

As for the Shark attacks, the bites drew blood and left scars... but the shark should never have been in my paddling pool.

Good we have no piranhas here ;-)

Is there anything left to do here?

Yes. The "Files are not copied on update" which still needs doing/testing/debugging/fixing

... if we will ever be able to reproduce it. As far as I understand, that seems to be not easy.

Im kind of hoping it was an opcache thing, but re-reading Nic's explanation Im now not sure. Like you said it seems hard to replicate, so my plan (Before I was sidetracked with he FTP Layer) was to write the code to compare each files checksum after installation and to DIE if the checksum was different (for now just DIE and report the findings, and not actually rewrite the file again like Nic's workaround does) - and publish that as a PR so people could hammer it on some sites, and see if we could then replicate it...

If we can then replicate it - by using the the PR that DIES if the hash of the files in the package are different to the hash of the file on the hard drive - then at least we have something to look at further. Even better if it only happens sometimes. Even better if its reproducible on demand.

DIES = Tell the user somehow.

That sounds like a good plan.