Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Won't Fix
-
10.1(EOL), 10.2(EOL), 10.3(EOL), 10.4(EOL)
-
None
Description
We have been building against system PCRE for some years in Debian now, using patch: https://salsa.debian.org/mariadb-team/mariadb-10.4/-/blob/master/debian/patches/0024-Revert-to-using-system-pcre-library.patch
In https://jira.mariadb.org/browse/MDEV-14900 Serg commented that we should use system PCRE but first find out why we were not using them in the first place.
To me it looks like we used the bundled PCRE simply because we lacked that option or because system PCRE gave errors:
- https://jira.mariadb.org/browse/MDEV-5620
- https://jira.mariadb.org/browse/MDEV-14640
- https://jira.mariadb.org/browse/MDEV-17868
Nowadays those problems should not exists anymore, as the use of the patch in downstream Debian packaging proves.
otto, it's clear from your 0024-Revert-to-using-system-pcre-library.patch. PCRE has a bug https://bugs.exim.org/show_bug.cgi?id=2173 that will not be fixed upstream. We have fixed it in our bundled pcre copy with the patch 440157cbbe7.
This check, that you removed in your 0024 patch, checks whether system pcre have this bug fixed. If yes — we use system pcre, if no — we use our bundled version. Some distros use our patch and then our cmake check selects system pcre. On other distros it selects bundled pcre.
You disabled this check — you got a broken server binary where regex sql function can crash the server.