Code review comment for lp:~jml/pkgme-devportal/uniqueness-constraint

On 22 August 2012 19:45, James Westby <email address hidden> wrote:
> Review: Needs Information
>
> Hi,
>
> 1. If this is going in the patch-00001.sql file then we want to land it before applying that patch?

Yes. I don't think it really matters though, as could easily make this
patch-00002.

> 2. What makes you think this isn't a good idea.

See below.

> 3. Perhaps we want a XFAIL test for now?

I'm not sure, tbh.

> 4. In fact I'm not sure why the test passes? Is NULL always considered unique?

Yes. That's what got me. Specifically (a, b, c, NULL) is considered
not equal to (a, b, c, NULL) for the purposes of uniqueness, even
though (a, b, c) _is_ considered equal to (a, b, c).

I don't currently see a path forward that doesn't involve potential
data corruption or downtime.

jml