Regenerate patch files on cache clear

[mattwire]

1. If the patch doesn't exist we generate it.
2. If the core file is updated we regenerate it.
3. If the extension that requires patchwork is updated we don't regenerate patches.

This "fixes" 3 so that a cache clear in CiviCRM will force regeneration of patch files. This is useful because it ensures that the client site is using the latest version of patched files if you issue them with an updated extension as they are usually familiar with "clearing caches". The alternative is to ask them to manually delete the contents of the patchwork directory which may or may not be easy...

[artfulrobot]

extension updates should be covered anyway, as it goes on mtime of the file. I agree a 'regenerate now' button would be nice, but I think adding to the already slow clear cache thing is a bit much.

Is it that you're un-tarring extensions and therefore they are retaining original mtimes?

[mattwire]

Is it that you're un-tarring extensions and therefore they are retaining original mtimes?

patchwork is not checking the mtimes of extension files - only core files. So if you update the extension but not core it doesn't detect changes.

[artfulrobot]

I think I need to understand the use case better. Are you using patchwork to patch other extensions' files? (does that even work?)

😕

[mattwire]

No :-) Simple / standard use-case.

This is the extension: https://github.com/mjwconsult/civicrm-2wish
It configures patchwork to patch 3 core files. But if you update that extension above (eg. because you add a new feature) and you don't update CiviCRM core the patch files won't regenerate unless you delete them manually.

[mattwire]

Maybe it's meant to pick up the mtime of the overridden core files rather than the original core files?

[artfulrobot]

Ah no, I get you.

In which case, I'd prefer an option whereby we provide a 'regenerate all patches' (or possibly a list of them) API action, and then call that in an extension's upgrader.

[mattwire]

Ok, that sounds good. If you don't mind I'll leave this open until I have refactored per above.

[artfulrobot]

Closing in favour of #3