> On Jan 28, 2026, at 14:14, Chao Li <li.evan.chao@gmail.com> wrote:
>
> Hi Hackers,
>
> While working on the patch [1], I was looking at the handling of AT_AddIndexConstraint in ATPrepCmd():
> ```
> ATPrepCmd()
> {
> case AT_AddIndexConstraint: /* ADD CONSTRAINT USING INDEX */
> ATSimplePermissions(cmd->subtype, rel, ATT_TABLE | ATT_PARTITIONED_TABLE);
> /* This command never recurses */
> /* No command-specific prep needed */
> pass = AT_PASS_ADD_INDEXCONSTR;
> }
> ```
>
> I initially thought the comment about “never recurses” was stale, but after some debugging, I found that this branch
isactually unreachable. So leaving the code and comments in an unreachable branch would be confusing for readers.
>
> This patch cleans up the handling by putting an Assert(false) there and adding a comment to explain why this code
pathis unreachable. I did think about just deleting the branch, but decided to keep it: if it were removed entirely,
readersmight wonder why AT_AddIndexConstraint is not handled in ATPrepCmd() and end up spending time debugging this
themselves.
>
I thought over and decided to delete AT_AddIndexConstraint from ATPrepCmd, which should be cleaner.
PFA v2:
* Deleted AT_AddIndexConstraint from ATPrepCmd
* Added a comment under AT_AddConstraint to explain how AT_AddIndexConstraint is handled
Best regards,
--
Chao Li (Evan)
HighGo Software Co., Ltd.
https://www.highgo.com/
