Horton

Thread

Commits

  1. Fix overly large values/nulls arrays

  1. BUG #18755: Using Natts_pg_trigger rather than Natts_pg_event_trigger for event trigger

    The Post Office <noreply@postgresql.org> — 2024-12-29T08:52:31Z

    The following bug has been logged on the website:

Bug reference:      18755
Logged by:          Xin Zhang
Email address:      zhanghien@qq.com
PostgreSQL version: 17.2
Operating system:   Any
Description:        

In function insert_event_trigger_tuple, the datum array is declared with
length Natts_pg_trigger:
```c
	Datum		values[Natts_pg_trigger];
	bool		nulls[Natts_pg_trigger];
```

Natts_pg_trigger should only be used in pg_trigger tuple and
Natts_pg_event_trigger is for pg_event_trigger?
I think this could be a typo. Maybe the author of function
insert_event_trigger_tuple copied the variables from CreateTriggerFiringOn
but forgot to change the array length to Natts_pg_event_trigger.

The current code works because Natts_pg_trigger is 19 and
Natts_pg_event_trigger is 7, the pg_event_trigger values array is allocated
with some extra items which is never accessed with index >= 7.

  2. Re: BUG #18755: Using Natts_pg_trigger rather than Natts_pg_event_trigger for event trigger

    张鑫 <zhanghien@qq.com> — 2024-12-29T09:37:18Z

    Hi&nbsp;everyone!


Here is the fix for BUG #18755. I fail to upload the patch with my mailbox and I just paste the text here because it's simple enough.


From 94009c0ae26443d03bd81ec10a1b666a4c10def4 Mon Sep 17 00:00:00 2001
From: "Xin Zhang" <zhanghien@qq.com&gt;
Date: Sun, 29 Dec 2024 17:08:53 +0800
Subject: [PATCH] Replace Natts_pg_trigger with Natts_pg_event_trigger for
&nbsp;event trigger


---
&nbsp;src/backend/commands/event_trigger.c | 4 ++--
&nbsp;1 file changed, 2 insertions(+), 2 deletions(-)


diff --git a/src/backend/commands/event_trigger.c b/src/backend/commands/event_trigger.c
index dcfc1dbaffd..1f5872554d4 100644
--- a/src/backend/commands/event_trigger.c
+++ b/src/backend/commands/event_trigger.c
@@ -276,8 +276,8 @@ insert_event_trigger_tuple(const char *trigname, const char *eventname, Oid evtO
&nbsp;	Relation	tgrel;
&nbsp;	Oid			trigoid;
&nbsp;	HeapTuple	tuple;
-	Datum		values[Natts_pg_trigger];
-	bool		nulls[Natts_pg_trigger];
+	Datum		values[Natts_pg_event_trigger];
+	bool		nulls[Natts_pg_event_trigger];
&nbsp;	NameData	evtnamedata,
&nbsp;				evteventdata;
&nbsp;	ObjectAddress myself,
--&nbsp;
2.39.3

  3. Re: BUG #18755: Using Natts_pg_trigger rather than Natts_pg_event_trigger for event trigger

    David Rowley <dgrowleyml@gmail.com> — 2024-12-29T10:36:40Z

    On Sun, 29 Dec 2024 at 22:37, 张鑫 <zhanghien@qq.com> wrote:
> Here is the fix for BUG #18755. I fail to upload the patch with my mailbox and I just paste the text here because it's simple enough.

I've not seen this bug report. Perhaps it's caught up in moderation?

> --- a/src/backend/commands/event_trigger.c
> +++ b/src/backend/commands/event_trigger.c
> @@ -276,8 +276,8 @@ insert_event_trigger_tuple(const char *trigname, const char *eventname, Oid evtO
>   Relation tgrel;
>   Oid trigoid;
>   HeapTuple tuple;
> - Datum values[Natts_pg_trigger];
> - bool nulls[Natts_pg_trigger];
> + Datum values[Natts_pg_event_trigger];
> + bool nulls[Natts_pg_event_trigger];
>   NameData evtnamedata,
>   evteventdata;
>   ObjectAddress myself,

Yes, that's better. However, no particular harm done since pg_trigger
has more columns than pg_event_trigger.  For that reason, it's
probably fine to fix in master only. The number of columns in those
tables in the back branches isn't going to change.

I scanned around to see if there's been any similar copy/paste errors,
but couldn't see any.

I'll take care of pushing this. Thanks for the patch.

David