Fix rice-box not found problem

[morgo]

This fixes #5654 after it was discovered that it does not work as intended when the config directory is not present!

Rice boxes do not support being called in an Init function, which means that without refactoring we can currently not use this feature for the init_db.sql file.

Unfortunately this means that the examples need to find a path to init_db.sql. I am not really happy with the solution I went with, but I copied the config/init_db.sql file to examples/local.

The alternative is that we could re-introduce VTROOT as a dependency, or add some sort of symlink in the repo + a copy on making packages. As I said - I'm not happy with the current solution, and really open to ideas.

Thanks @enisoc : it now works as expected.

Edit: This has a second behavior change, which is that the embedded config is no longer rebuilt on make build. This created potentially noisy commits, since it kept bumping the timestamps. It also made the automated build system think it was working on dirty trees.

Signed-off-by: Morgan Tocker tocker@gmail.com

[deepthi]

I think it is actually a good thing to have the example be more self-contained and less reliant on files spread out in the code base.

[enisoc]

Rice boxes do not support being called in an Init function, which means that without refactoring we can currently not use this feature for the init_db.sql file.

We wouldn't need to call it from an init() function here. This is just a regular function that happens to be called Init(). It gets called explicitly after main() starts, so it should be fine to call rice.FindBox() from inside Mysqld.Init(). We just can't call rice.FindBox() from a package-level variable initializer as we did previously.

[morgo]

We wouldn't need to call it from an init() function here. This is just a regular function that happens to be called Init(). It gets called explicitly after main() starts, so it should be fine to call rice.FindBox() from inside Mysqld.Init(). We just can't call rice.FindBox() from a package-level variable initializer as we did previously.

You are right! it is an Init() function, not an init() function. I've updated it, and it seems to work locally. Let's see if CI works.

I think it is actually a good thing to have the example be more self-contained and less reliant on files spread out in the code base.

That's the goal. I just don't didn't like the implementation.