[PATCH] snowball: fix potential NULL dereference

=?utf-8?q?=D0=9A=D0=BE=D1=80=D0=BE=D1=82=D0=BA=D0=BE=D0=B2_=D0=9C=D0=B0=D0=BA=D1=81=D0=B8=D0=BC?= <m.korotkov@postgrespro.ru> writes:

I found the case of potential NULL pointer dereference.
In snowball/libstemmer/api.c if we transfer control to the SN_create_env() function
by using the error label when there is a memory allocation error of z->p or z->S,
we can then dereference the NULL pointer z->S in the function SN_close_env().
Added the pointer check for avoiding a potential problem.

I believe you are right: OOM partway through SN_create_env would fail.
However, backend/snowball is not our code so applying our own local
patch is not the way to fix it. You should report this upstream;
see src/backend/snowball/README.

(Whenever they apply the patch, we should then re-sync...)

regards, tom lane

Hi
On 2025-02-11 19:53, Tom Lane wrote:

You should report this upstream;
(Whenever they apply the patch, we should then re-sync...)

FYI, the patch to fix this problem was applied by upstream.
---
Best regards, Korotkov Maksim
PostgresPro
m.korotkov@postgrespro.ru

Maksim Korotkov <m.korotkov@postgrespro.ru> writes:

On 2025-02-11 19:53, Tom Lane wrote:

You should report this upstream;
(Whenever they apply the patch, we should then re-sync...)

FYI, the patch to fix this problem was applied by upstream.

Thanks. Barring objections, I'll do a full resync per
src/backend/snowball/README on HEAD, and then extract just
the null-dereference fix to apply to our back branches.

regards, tom lane