mariadb-10.3 crashes on macOS Monterey

[GreggGreen (Gregg Green)]

Server starts but any query causes the server to crash and restart. Using brew until this is fixed.

[GreggGreen (Gregg Green)]

Error Log

[ryandesign (Ryan Carsten Schmidt)]

Do newer versions work? Newest we have in MacPorts is mariadb-10.7. If you see the problem with the latest version, you should report it to the developers:

To report this bug, see ​https://mariadb.com/kb/en/reporting-bugs

[GreggGreen (Gregg Green)]

I'll go up the list of mariadb versions to see if any work and let you know.

[GreggGreen (Gregg Green)]

10.4 has same crashing issue as 10.3

Here are the build messages:

---> Computing dependencies for mariadb-10.4 ---> Fetching archive for mariadb-10.4 ---> Attempting to fetch mariadb-10.4-10.4.22_0.darwin_21.x86_64.tbz2 from ​https://packages.macports.org/mariadb-10.4 ---> Attempting to fetch mariadb-10.4-10.4.22_0.darwin_21.x86_64.tbz2 from ​https://ywg.ca.packages.macports.org/mirror/macports/packages/mariadb-10.4 ---> Attempting to fetch mariadb-10.4-10.4.22_0.darwin_21.x86_64.tbz2 from ​https://mse.uk.packages.macports.org/mariadb-10.4 ---> Fetching distfiles for mariadb-10.4 ---> Attempting to fetch server-10.4.22.tar.gz from ​https://distfiles.macports.org/mariadb-10.4 ---> Attempting to fetch mariadb-connector-c-735a7299dbae19cc2b82b9697becaf90e9b43047.tar.gz from ​https://distfiles.macports.org/mariadb-10.4 ---> Attempting to fetch wsrep-lib-22921e7082ddfb45222f21a585aa8b877e62aa8b.tar.gz from ​https://distfiles.macports.org/mariadb-10.4 ---> Attempting to fetch wsrep-API-76cf223c690845bbf561cb820a46e06a18ad80d1.tar.gz from ​https://distfiles.macports.org/mariadb-10.4 ---> Verifying checksums for mariadb-10.4 ---> Extracting mariadb-10.4 ---> Applying patches to mariadb-10.4 ---> Configuring mariadb-10.4 Warning: Configuration logfiles contain indications of -Wimplicit-function-declaration; check that features were not accidentally disabled:

select: found in build/CMakeFiles/CMakeError.log

---> Building mariadb-10.4 ---> Staging mariadb-10.4 into destroot ---> Installing mariadb-10.4 @10.4.22_0 The mariadb-10.4 client has been installed. To install the mariadb-10.4 server, install the mariadb-10.4-server port. ---> Activating mariadb-10.4 @10.4.22_0 ---> Cleaning mariadb-10.4 ---> Scanning binaries for linking errors ---> No broken files found. ---> No broken ports found. ---> Some of the ports you installed have notes:

mariadb-10.4 has the following notes:

On activation if no /opt/local/etc/mariadb-10.4/my.cnf file exists one will be created which loads /opt/local/etc/mariadb-10.4/macports-default.cnf.

If a /opt/local/etc/mariadb-10.4/my.cnf file exists MacPorts does not touch it and any changes you make to /opt/local/etc/mariadb-10.4/my.cnf will be preserved (e.g., during port upgrades, deactivations or activations). /opt/local/etc/mariadb-10.4/my.cnf is a good place to customize your mariadb-10.4 installation.

Any changes made to /opt/local/etc/mariadb-10.4/macports-default.cnf will be lost during port upgrades, deactivations or activations so you are advised to not make changes here. Currently /opt/local/etc/mariadb-10.4/macports-default.cnf contains only one directive; to disable networking. With disabled networking it is possible to install and have running all the MacPorts mysql ports simultaneously.

[GreggGreen (Gregg Green)]

10.5 gets this message while building:

Warning: Configuration logfiles contain indications of -Wimplicit-function-declaration; check that features were not accidentally disabled:

select: found in build/CMakeFiles/CMakeError.log getthrid: found in build/CMakeFiles/CMakeError.log

[chrstphrchvz (Christopher Chavez)]

MariaDB not supporting and even crashing when built with OpenSSL 3 is a known issue: ​https://jira.mariadb.org/browse/MDEV-26771, ​https://jira.mariadb.org/browse/MDEV-25785.

MariaDB hardcodes the sizes of various OpenSSL data structures—even though OpenSSL 1.1 and later want programs to treat them as opaque and have no knowledge of their size—because MariaDB insists on allocating them on the stack for performance. There is a runtime compatibility check in MariaDB which should abort if the hardcoded sizes are insufficient; so if MariaDB is crashing rather than aborting, I think that may indicate their runtime compatibility check isn’t running properly. The stack traces show the crash occurring somewhere OpenSSL compatibility is assumed, so I wonder if that is being executed before the compatibility check.