On 2025/06/25 22:36, Álvaro Herrera wrote:
On 2025-Jun-25, Álvaro Herrera wrote:
Yeah, I think in this case we need to extract the constraint name so
that we have it available to print the COMMENT command, rather than
making any assumptions about it. In fact I suspect this would fail if
the table or column names are very long. For the other pg_dump uses of
this logic it doesn't matter AFAIR, but here I think we must be
stricter.
As attached.
Thanks for the patch! I agree with the approach, i.e., printing the not-null
constraint name only when there's a comment on it.
However, with the patch applied, I encountered a segmentation fault in pg_dump
as follows:
$ psql <<EOF
create table t (i int);
alter table t add constraint t_i_not_null not null i not valid;
comment on constraint t_i_not_null ON t IS 'iii';
EOF
$ pg_dump
Segmentation fault: 11
I'm bothered by this not having any tests -- I'll see about adding some
after lunch.
+1. Thanks!
Regards,
--
Fujii Masao
NTT DATA Japan Corporation
