gcc-15 fixes - use void for sig_handler returns

[heitbaum]

• The Jira issue number for this PR is: MDEV-35607

Description

fixes compile issue with gcc-15

../mysys/stacktrace.c: In function 'default_handle_fatal_signal':
../mysys/stacktrace.c:53:3: error: 'return' with no value, in function 
returning non-void [-Wreturn-mismatch]
    53 |   return;
       |   ^~~~~~
...

Release Notes

Fixes build with gcc-15 when combined with fixes from:

• https://jira.mariadb.org/browse/MDEV-35605
• https://jira.mariadb.org/browse/MDEV-35606

How can this PR be tested?

Testing can be performed using gcc-15 as a compiler

Basing the PR against the correct MariaDB version

• This is a new feature or a refactoring, and the PR is based against the main branch.
• This is a bug fix, and the PR is based against the earliest maintained branch in which the bug can be reproduced.

PR quality check

• I checked the CODING_STANDARDS.md file and my PR conforms to this where appropriate.
• For any trivial modifications to the PR, I am ok with the reviewer making the changes themselves.

[CLAassistant]

All committers have signed the CLA.

[grooverdan]

Interesting - note it looks like it was intended to autodetect this -

server/configure.cmake

Line 678 in f5821aa

CHECK_C_SOURCE_COMPILES("

- I haven't looked into the history as to why.

[heitbaum]

Interesting - note it looks like it was intended to autodetect this -

server/configure.cmake

Line 678 in f5821aa

CHECK_C_SOURCE_COMPILES("

• I haven't looked into the history as to why.

Yes. I'll have a look why once I get all the parts cleanly compiling. I thought it was going to be a quick one liner... :-(

[dr-m]

Interesting - note it looks like it was intended to autodetect this -

server/configure.cmake

Line 678 in f5821aa

CHECK_C_SOURCE_COMPILES("

• I haven't looked into the history as to why.

If that auto-detection is obsolete by now, maybe we could remove it and thereby reduce the CMake execution time. man 2 signal on my system mentions C11 (ISO/IEC 9899:2011). In https://en.cppreference.com/w/c/program/signal there is no mention of signal() returning void; it always was supposed to return a function pointer. I would expect that any currently supported compiler would support at least C11.

[heitbaum]

Interesting - note it looks like it was intended to autodetect this -

server/configure.cmake

Line 678 in f5821aa

CHECK_C_SOURCE_COMPILES("

• I haven't looked into the history as to why.

If that auto-detection is obsolete by now, maybe we could remove it and thereby reduce the CMake execution time. man 2 signal on my system mentions C11 (ISO/IEC 9899:2011). In https://en.cppreference.com/w/c/program/signal there is no mention of signal() returning void; it always was supposed to return a function pointer. I would expect that any currently supported compiler would support at least C11.

That article concurs with the changes I have made:

void (*signal( int sig, void (*handler) (int))) (int);
                        ^^^^

In that the handler is suppose to be void not sig_handler which happens to be int in this instance.

I have updated the commit messages to reference sig_handler so as to be more correct.

[vuvova]

as @grooverdan wrote, we need to understand why the check isn't working before we decide to remove or ignore it.
Here's how you can find it out:

• run cmake with whatever options you normally do it
• then run cmake -USIGNAL_RETURN_TYPE_IS_VOID --debug-trycompile
• then cd CMakeFiles/CMakeScratch/TryCompile-* && make

the last line might look differently depending on your cmake version and your make tool