proposal: make the build.zig.ini file a strict subset of TOML

[andrewrk]

Extracted from #14265.

Currently the build.zig.ini file is not valid TOML. Example:

[package]
name=libffmpeg
version=5.1.2

[dependency]
name=libz
url=https://github.com/andrewrk/libz/archive/f0e53cc2391741034b144a2c2076ed8a9937b29b.tar.gz
hash=c9b30cffc40999d2c078ff350cbcee642970a224fe123c756d0892f876cf1aae

[dependency]
name=libmp3lame
url=https://github.com/andrewrk/libmp3lame/archive/497568e670bfeb14ab6ef47fb6459a2251358e43.tar.gz
hash=9ba4f49895b174a3f918d489238acbc146bd393575062b2e3be33488b688e36f

To make it valid TOML:

[package]
name="libffmpeg"
version="5.1.2"

[[dependency]]
name="libz"
url="https://github.com/andrewrk/libz/archive/f0e53cc2391741034b144a2c2076ed8a9937b29b.tar.gz"
hash="c9b30cffc40999d2c078ff350cbcee642970a224fe123c756d0892f876cf1aae"

[[dependency]]
name="libmp3lame"
url="https://github.com/andrewrk/libmp3lame/archive/497568e670bfeb14ab6ef47fb6459a2251358e43.tar.gz"
hash="9ba4f49895b174a3f918d489238acbc146bd393575062b2e3be33488b688e36f"

It also should be renamed with the .toml extension, and furthermore, [package] should be renamed to [project] to match the correct terminology.

Let's have a short list of accepted TOML features and only use those.

See competing proposal:

• proposal: instead of build.zig.ini, use build.zig.zon #14290

I don't like this. People might want to use their own TOML generators for tooling, but then they have to make sure they only generate a subset, which could itself be a hassle to do with their tools as well as causing confusion around the standard. In the worst case we splinter the TOML standard.

[thejoshwolfe]

it does seem dishonest to name it .toml if it isn't really TOML. maybe a different extension would resolve that though.

After reading the spec, it seems like maybe we should use 'single quote' strings instead of "double quote" strings? To @EleanorNB's point, there are an awful lot of ways to express the same thing in TOML. Restricting to a subset solves that problem for the compiler's parser, but pushes the problem onto everyone's generator.

I've just got to complain about this line from the TOML spec: https://toml.io/en/v1.0.0

Any reference to an array of tables points to the most recently defined table element of the array.

TOML is a mistake. This is way too complex and confusing. What even is the O in the name. Woe is humanity. If only JSON had allowed comments and trailing commas; what a fantastic mess we could have avoided.