299230 - Schema-Modification Functions Shouldn't Die on A DB-side Deletion Failure

Status: RESOLVED FIXED

(Bugzilla :: Database, defect)

Description

[Max Kanat-Alexander]

Right now, when checksetup tries to "drop" anything that isn't actually in the
DB, it gets an error and dies.

Instead, it should just accept this error and carry on with its life.

Basically, this means that the "do" statements for a bz_drop function should be
inside an eval block. The error should still be printed, but we should delete
the object from the Schema anyhow.

Comment 1

[Max Kanat-Alexander]

I think there's a good enough chance of this causing a significant support issue
that it should go either into 2.20 or 2.20.1.

Comment 2

[Max Kanat-Alexander]

Attachment: Make deletion "nice"

OK, here it is. We still print a warning to the user that we couldn't do
something, but we are otherwise very "nice" about SQL failures during deletion.

Comment 3

[Max Kanat-Alexander]

Honestly, I wouldn't block the release on this (it could wait for 2.20.1), but I
think we'll get it in anyway.

Comment 4

[Marc Schumann [:Wurblzap]]

What other reasons, besides a local customization having removed the item in
question already, are there for one of the statements to fail?

Comment 5

[Max Kanat-Alexander]

(In reply to comment #4)
> What other reasons, besides a local customization having removed the item in
> question already, are there for one of the statements to fail?

  No other reason. However, we should be at least somewhat nice to local
customizations, because it's nearly impossible for people to locally modify the
stored Schema object, right now.

Comment 6

[Marc Schumann [:Wurblzap]]

Comment on attachment 188884 [details] [diff] [review]
Make deletion "nice"

Hm. This seems to work, but I'm not entirely sure that this is really what we
want to do. I can't put my finger on it, though. Please ask somebody else for
review :)

Comment 7

[Max Kanat-Alexander]

Comment on attachment 188884 [details] [diff] [review]
Make deletion "nice"

I am really sure that this is what we want to do.

If we go to *delete* something, and we find it's already *not there*, then we
shouldn't care.

The only time we could run into trouble is if the SQL hits some error other
than the non-existence of the field. But that's pretty rare. (Much rarer than
the problem I'm trying to fix.)

Comment 8

[Joel Peshkin]

Comment on attachment 188884 [details] [diff] [review]
Make deletion "nice"

r=joel if you've tested it.

Comment 9

[Max Kanat-Alexander]

I have tested it.

Comment 10

[Max Kanat-Alexander]

Tip:

Checking in Bugzilla/DB.pm;
/cvsroot/mozilla/webtools/bugzilla/Bugzilla/DB.pm,v  <--  DB.pm
new revision: 1.65; previous revision: 1.64
done

2.20:

Checking in Bugzilla/DB.pm;
/cvsroot/mozilla/webtools/bugzilla/Bugzilla/DB.pm,v  <--  DB.pm
new revision: 1.56.2.3; previous revision: 1.56.2.2
done