Fix MySQL 7.4 vs 8 compatibility bugs

While I do appreciate the effort to improve the logic, I really think this should be a simpler solution to select between old (<=8.03) vs new (>=8.0.4).

An alternative solution - add a setting that controls whether to use the old method or the new method.

Still to do - hook_install() should have some logic to determine the best option, and it'd might be worth having a hook_requirements to compare this logic with what is defined in the settings.

The last submitted patch, 25: scanner-n3316862-25.patch, failed testing. View results →

Assigning it to myself while I finish off this idea..

This adds a setting for controlling it, and adds a hook_requirements item for checking it. I tried to add a hook_install() bit to do the same logic, but it was failing, so that still needs to be done.

Alternatively we could rework the logic to just use the WordBoundariesHelper::shouldBe() method to decide at runtime.

I might expand it to have an "auto" option, which would be the default, and then let site admins change it if they specifically want to.

The last submitted patch, 28: scanner-n3316862-28.patch, failed testing. View results →
- codesniffer_fixes.patch Interdiff of automated coding standards fixes only.

How about this? It defaults to "auto", but it can be overridden if needed.

The last submitted patch, 31: scanner-n3316862-31.patch, failed testing. View results →
- codesniffer_fixes.patch Interdiff of automated coding standards fixes only.

This should fix the error from the last patch.

The last submitted patch, 33: scanner-n3316862-33.patch, failed testing. View results →
- codesniffer_fixes.patch Interdiff of automated coding standards fixes only.

This resolves the problem locally, with MariaDB anyways.

The last submitted patch, 35: scanner-n3316862-35.patch, failed testing. View results →
- codesniffer_fixes.patch Interdiff of automated coding standards fixes only.

Ok. Now, back to fixing that (existing?) error, and then testing it on MySQL 8 too.

Added an implementation which works off a PCRE flag instead of the REGEXP_LIKE function.

See https://jira.mariadb.org/browse/MDEV-4425

It was decided not to implement the REGEXP_LIKE function, because PCRE supports flags in the pattern syntax. There is no a need for an extra argument that would contain flags.

Example SQL fiddle showcasing support for the flag amongst database engines: https://dbfiddle.uk/bh8he4yd

@codebymikey: I really appreciate your work improving the patch, thank you!

Can you please confirm this contains all of the changes, and that the interdiff is accurate?

@DamienMcKenna, no worries, and yes I can confirm that the interdiff is accurate!

Committed. Thank you!

Cheers, just out of curiousity, is there a reason you commit locally via patches rather than merging through issue forks?

Also, are we fine to tag a new release sometime soon?

DamienMcKenna → closed merge request !22

DamienMcKenna → closed merge request !19

For now I prefer working with patches.

I'm going to tag a new RC with these fixes, and then there are a few other small bug fixes needed before tagging the stable release.

Automatically closed - issue fixed for 2 weeks with no activity.