Local build improvement

[NQNStudios]

I went through all of @clort81's challenges when building the latest source code, and added automation/documentation for everything with a known solution.

• Catch2 is now a subtree, not a submodule, because subtrees automatically clone without git clone --recursive or git submodule update --init
• Added TGUI 0.9 as a subtree, and updated the Scons build to build it from source using C++14 if the library isn't detected installed on the system.
• Updated README.md to explain the annoying particularities of installing TGUI if folks choose to do it themselves

[NQNStudios]

I tested it locally on a fresh clone and actually it needs a couple more things.

[NQNStudios]

Ok, I just tested it locally and a fresh clone on linux can just run scons as long as they have all the dependencies, plus cmake to build TGUI

[CelticMinstrel]

• Catch2 is now a subtree, not a submodule, because subtrees automatically clone without git clone --recursive or git submodule update --init

I'm a bit unsure about this part…

[NQNStudios]

It's basically the same as a submodule, but we'll never have to remind anyone to call submodule update --init. I pulled the same version of Catch2 as we were using before.

If Catch2 were an optional dependency I could see an advantage to not always cloning it. But currently we have a state where everyone who tries to build from a normal git clone will get an easily avoidable compiler error--I don't see the point.

[CelticMinstrel]

After reading up on subtree and checking why I made Catch a submodule in the first place, I'm even less sold on the idea of the subtree. If the only issue is that submodules aren't automatically cloned, I think I'd rather just add a step in scons that makes sure the submodules are present before building things that depend on them. In other words, have scons call git submodule update --init for you.

And I also notice that you made TGUI a subtree here. I think linking it in as a submodule would actually be a good idea. I guess that would suggest that scons then needs to call cmake for TGUI, which is really weird, but whatever.

[NQNStudios]

People building with Xcode or Visual Studio will still fall into the trap of Catch2 being missing, though

…

On Thu, May 30, 2024 at 5:46 PM Celtic Minstrel ***@***.***> wrote: After reading up on subtree and checking why I made Catch a submodule in the first place, I'm even less sold on the idea of the subtree. If the only issue is that submodules aren't automatically cloned, I think I'd rather just add a step in scons that makes sure the submodules are present before building things that depend on them. In other words, have scons call git submodule update --init for you. And I also notice that you made TGUI a subtree here. I think linking it in as a submodule would actually be a good idea. — Reply to this email directly, view it on GitHub <#356 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AATXBKMHSNXV7RNFR7MV3B3ZE6233AVCNFSM6AAAAABIRMYIL2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNBRGAYTCMBUGQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[CelticMinstrel]

That's true, but people building with Xcode or Visual Studio would still be able to build the game even if Catch is missing. They'd only get a build failure if they try to build the tests. (This should be true of scons too, I just haven't bothered to work out how to make scons have multiple targets.)