372974 - (bz-plugin-attach) Hook to process inbound attachments

Status: RESOLVED FIXED

(Bugzilla :: Attachments & Requests, enhancement)

Description

[Max Kanat-Alexander]

There are various enhancements requests in Bugzilla to "Do some crazy thing with this patch when people submit it."

We should just have a hook that can modify attachment data before it's put into the database.

Once we have Attachment::create, this should be pretty easy. The Attachment->create() bug should block this one.

Comment 1

[Frédéric Buclin]

I don't think bugs marked as depending on this one should be implemented upstream, because validating testcases (bug 63580), validating patches for compilation errors (bug 91546), rejecting unwanted diff formats (bug 119502), testing patches thanks to a test suite (bug 167056), or scanning attachments for viruses (bug 317013) are all stuff which is external to Bugzilla. We would need a by far too complex code to handle all these possibilities (code repository, test suite, antivirus software spec, ...). I indeed think fixing this bug and adding some hooks to let people implement external tools is the way to go. That's why I plan to mark all the depending bugs as WONTFIX. Any objection, Max?

Comment 2

[Max Kanat-Alexander]

Well, until we have a method for marking things as being good plugin ideas, let's keep them open. Once we have some keyword or agreed-on method of saying, "this should be a plugin, not in core" then we can close them. Otherwise it's hard to differentiate them from other WONTFIX bugs that are actually bad ideas.

Comment 3

[Frédéric Buclin]

(In reply to comment #2)
> hard to differentiate them from other WONTFIX bugs that are actually bad ideas.

So we should discuss a way to exclude these "should be implemented as plugins" bugs from our buglists. This would also avoid having people working on them and then crying that we don't want their patches upstream.

Comment 5

[Vitaly Fedrushkov]

> > hard to differentiate them from other WONTFIX bugs that are actually bad ideas.

> So we should discuss a way to exclude these "should be implemented as plugins"

Looks like we need a tracking metabug for would-be extensions

Comment 6

[Max Kanat-Alexander]

Fixed by bug 480986.