Re: Enables to call Unregister*XactCallback() in Call*XactCallback()

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Andres Freund <andres(at)anarazel(dot)de>
Cc: Hao Wu <gfphoenix78(at)gmail(dot)com>, pgsql-hackers(at)lists(dot)postgresql(dot)org
Subject: Re: Enables to call Unregister*XactCallback() in Call*XactCallback()
Date: 2022-09-26 22:05:34
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

Andres Freund <andres(at)anarazel(dot)de> writes:
> On 2022-03-29 14:48:54 +0800, Hao Wu wrote:
>> It's a natural requirement to unregister the callback for transaction or
>> subtransaction when the callback is invoked, so we don't have to
>> unregister the callback somewhere.

> You normally shouldn'd need to do this frequently - what's your use case?
> UnregisterXactCallback() is O(N), so workloads registering / unregistering a
> lot of callbacks would be problematic.

It'd only be slow if you had a lot of distinct callbacks registered
at the same time, which doesn't sound like a common situation.

>> Luckily, we just need a few lines of code to support this feature,
>> by saving the next pointer before calling the callback.

> That seems reasonable...

Yeah. Whether it's efficient or not, seems like it should *work*.
I'm a bit inclined to call this a bug-fix and backpatch it.

I went looking for other occurrences of this code in places that have
an unregister function, and found one in ResourceOwnerReleaseInternal,
so I think we should fix that too. Also, a comment seems advisable;
that leads me to the attached v2.

regards, tom lane

Attachment Content-Type Size
safely-delete-callback-2.patch text/x-diff 1.9 KB

In response to


Browse pgsql-hackers by date

  From Date Subject
Next Message Nathan Bossart 2022-09-26 22:13:39 Re: Enables to call Unregister*XactCallback() in Call*XactCallback()
Previous Message David Rowley 2022-09-26 21:59:04 Re: First draft of the PG 15 release notes