public inbox for [email protected]
help / color / mirror / Atom feedReplace PROC_QUEUE / SHM_QUEUE with ilist.h
2+ messages / 1 participants
[nested] [flat]
* Replace PROC_QUEUE / SHM_QUEUE with ilist.h
@ 2022-11-20 05:59 Andres Freund <[email protected]>
0 siblings, 1 reply; 2+ messages in thread
From: Andres Freund @ 2022-11-20 05:59 UTC (permalink / raw)
To: pgsql-hackers; +Cc: Robert Haas <[email protected]>; Thomas Munro <[email protected]>; Tom Lane <[email protected]>; Amit Langote <[email protected]>
Hi,
In [1] Robert justifiably complained about the use of PROC_QUEUE. I've
previously been bothered by this in [2], but didn't get around to finishing
the patches.
One thing that made me hesitate was the naming of the function for a) checking
whether a dlist_node is a list, b) initializing and deleting nodes in a way to
allow for a). My patch adds dlist_node_init(), dlist_delete_thoroughly() /
dlist_delete_from_thoroughly() / dclist_delete_from_thoroughly() and
dlist_node_is_detached(). Thomas proposed dlist_delete_and_reinit() and
dlist_node_is_linked() instead.
Attached is a revised version of the patches from [2].
I left the naming of the detached / thoroughly as it was before, for
now. Another alternative could be to try to just get rid of needing the
detached state at all, although that likely would make the code changes
bigger.
I've switched the PROC_QUEUE uses to dclist, which we only got recently. The
prior version of the patchset contained a patch to remove the use of the size
field of PROC_QUEUE, as it's only needed in a few places. But it seems easier
to just replace it with dclist for now.
Robert had previously complained about the ilist.h patch constifying some
functions. I don't really understand the complaint in this case - none of the
cases should require constifying outside code. It just allows to replace
things like SHMQueueEmpty() which were const, because there's a few places
that get passed a const PGPROC. There's more that could be constified
(removing the need for one unconstify() in predicate.c - but that seems like a
lot more work with less clear benefit. Either way, I split the constification
into a separate patch.
Comments?
Greetings,
Andres Freund
[1] https://www.postgresql.org/message-id/20221117201304.vemjfsxaizmt3zbb%40awork3.anarazel.de
[2] https://www.postgresql.org/message-id/20200211042229.msv23badgqljrdg2%40alap3.anarazel.de
^ permalink raw reply [nested|flat] 2+ messages in thread
* Re: Replace PROC_QUEUE / SHM_QUEUE with ilist.h
@ 2022-12-03 18:17 Andres Freund <[email protected]>
parent: Andres Freund <[email protected]>
0 siblings, 0 replies; 2+ messages in thread
From: Andres Freund @ 2022-12-03 18:17 UTC (permalink / raw)
To: pgsql-hackers; +Cc: Robert Haas <[email protected]>; Thomas Munro <[email protected]>; Tom Lane <[email protected]>; Amit Langote <[email protected]>
Hi,
On 2022-11-19 21:59:30 -0800, Andres Freund wrote:
> In [1] Robert justifiably complained about the use of PROC_QUEUE. I've
> previously been bothered by this in [2], but didn't get around to finishing
> the patches.
>
> One thing that made me hesitate was the naming of the function for a) checking
> whether a dlist_node is a list, b) initializing and deleting nodes in a way to
> allow for a). My patch adds dlist_node_init(), dlist_delete_thoroughly() /
> dlist_delete_from_thoroughly() / dclist_delete_from_thoroughly() and
> dlist_node_is_detached(). Thomas proposed dlist_delete_and_reinit() and
> dlist_node_is_linked() instead.
Any comments on these names? Otherwise I'm planning to push ahead with the
names as is, lest I forget this patchset for another ~2 years.
Greetings,
Andres Freund
^ permalink raw reply [nested|flat] 2+ messages in thread
end of thread, other threads:[~2022-12-03 18:17 UTC | newest]
Thread overview: 2+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2022-11-20 05:59 Replace PROC_QUEUE / SHM_QUEUE with ilist.h Andres Freund <[email protected]>
2022-12-03 18:17 ` Andres Freund <[email protected]>
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox