Great minds think alike: T76245: Create ability to trivially spin up MediaWiki instance of a particular patch/diff: patchdemo tool.

of course, it might not be enough to teach MW this trick - in order to accomplish this you might need to teach gerrit to provide the "raw + patch",, and even do it in the correct mime-type (just like mw itself knows how to do when you tell it

?title=x&action=raw&ctype=whatever

if gerrit does not know how to do that, the gerrit team will probably be willing to teach it - the ability to do that can be useful elsewhere too.

peace.

In T78476#846276, @matmarex wrote:

Great minds think alike: T76245: Create ability to trivially spin up MediaWiki instance of a particular patch/diff: patchdemo tool.

i think the main difference here is that my suggestion has nothing to do with "spinning off" a whole new installation (locally or on a server). i am talking *only* about feeding the user's browser modified sources - MW does not even have to manage those sources - this is the job of gerrit (or git, if the patch is already checked-in to some branch).

T76245 is about "make it easier for developers to test a patch". my suggestion is about "make it _possible_ for a non-developer to test a patch"
(with known limitation that this can only work for purely JS/css patches).

peace.

Code for getting URL of script based on specific changeset can also use gitblit:

// example: getPatchsetRaw('mediawiki/core', 'a700f47e068d9eae0afb5efa6bac65ae0ef925e7', 'resources/mediawiki.special/mediawiki.special.createAccount.js')
function getPatchsetRaw(module, patchset, file) {
	return 'https://git.wikimedia.org/raw/'+encodeURIComponent(module)+'/'+patchset+'/'+encodeURIComponent(file);
}

Issues need to be considered:

1. One have to know all the associated files of a patch to load it
2. cross origin requests?
3. it may be complicated to handle dependencies of mw.loader (some dependencies may be loaded before the "fix" is loaded?)

How is this related to the MediaWiki-Javascript project?

Does this require changes in the MW core codebase itself?

As long as more or less everybody can propose a patch, this would be an obvious XSS-vector.