testsymbols: Use the existing symbol list in SDL_dynapi_procs.h

[icculus]

Rather than have another place we generate this list, let's use the one we already have!

This saves problems on older PRs that predate testsymbols, like #13264, and also prevents little mistakes, like SDL_SetX11EventHook being in the list twice.

[madebr]

The reason for testsymbols to exist (existed?) is (was?) to make sure non-dynapi ports have all symbols available.
We recently saw an error with ngage missing an implementation which would've been caught earlier by testsymbols.

[icculus]

It still does that, though...? It just gets the list of symbols from a different place, but it will work whether the port uses dynapi or not.

[icculus]

(If I'm wrong, though, let's fix this and/or revert it.)

[madebr]

It still does that, though...? It just gets the list of symbols from a different place, but it will work whether the port uses dynapi or not.

dynapi will totally complain about missing symbols for the (desktop-only) platforms that use it
This was the (stop-gap) fix for N-Gage: SDL_TimeToDateTime is a platform-specific function.

An alternative to testsymbols could be to make sure every SDL3 symbol is used at least once by a test.

[icculus]

Yes, but it's just using the dynapi symbol list to generate the same C code that was previously in testsymbols.c, so it should still catch things like the ngage problem. The intention is this program is exactly the same, it just doesn't manage this list of symbols separately.

I feel like you're saying something very clearly that I'm totally misunderstanding anyhow here. :)

[madebr]

I feel like you're saying something very clearly that I'm totally misunderstanding anyhow here. :)

Nah, you are correct. I misread the PR in the GitHub UI. I thought you removed testsymbols.c, but you didn't.
We're all good :)

Sorry about the misunderstanding!!