On 2021-Jun-08, Noah Misch wrote:
> On Wed, Jun 09, 2021 at 03:21:36PM +1200, Thomas Munro wrote:
> > On Wed, Jun 9, 2021 at 2:44 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > > ... or we could shorten those file names. I recall an episode
> > > awhile ago where somebody complained that their version of "tar"
> > > couldn't handle some of the path names in our tarball, so
> > > keeping things from getting to carpal-tunnel-inducing lengths
> > > does have its advantages.
Sure. I'm also the author of tuplelock-upgrade-no-deadlock -- see
commit de87a084c0a5. (Oleksii submitted it as "rowlock-upgrade-deadlock").
We could rename that one too while at it.
> > On Wed, Jun 9, 2021 at 2:51 PM Noah Misch <noah@leadboat.com> wrote:
> > > Not bad, but I would instead shorten the names to detach-[1234] or
> > > detach-partition-[1234]. The marginal value of the second word is low, and
> > > the third word helps even less.
>
> Better still, the numbers can change to something descriptive:
>
> detach-1 => detach-visibility
> detach-2 => detach-fk-FOO
> detach-3 => detach-incomplete
> detach-4 => detach-fk-BAR
>
> I don't grasp the difference between -2 and -4 enough to suggest concrete FOO
> and BAR words.
Looking at -2, it looks like a very small subset of -4. I probably
wrote it first and failed to realize I could extend that one rather than
create -4. We could just delete it.
We also have partition-concurrent-attach.spec; what if we make
everything a consistent set? We could have
partition-attach
partition-detach-visibility (-1)
partition-detach-incomplete (-3)
partition-detach-fk (-4)
--
Álvaro Herrera 39°49'30"S 73°17'W