rustc: Stricter matching of --cfg options

[dirk]

Fixes #31497.

[rust-highfive]

r? @brson

(rust_highfive has picked a reviewer for you, use r? to override)

[dirk]

A couple questions:

1. Is this the right way to approach this? Is comparing the strings after parsing too late, or should we make the parser function itself errors if there's any input string left after the parsed Ident?
2. How should tests of this behavior be done?

[alexcrichton]

I think that this still wants to parse a meta instead of an ident to we can still pass, for example, --cfg 'feature = "foo"'.

For tests you can add some compile-fail tests using the // compile-flags feature to help pass arguments down to rustc (invalid ones), and then you can use // error-pattern to match the output

[dirk]

@alexcrichton: If that's the case, then would it make sense to do some basic validation (perhaps with a simple visitor pattern) of the tree produced by the meta parser?

[alexcrichton]

It may make the most sense to just make sure there are no more tokens in the parser once a meta item is parsed (although I'm not sure how to do this). I think what's happening here is we're successfully parsing a meta item and we're just trashing the remaining tokens (when they should be considered invalid)

[dirk]

Oooh, good point! I'll look into that.

[dirk]

@alexcrichton: Comments addressed and compile-fail test added!

[alexcrichton]

@bors: r+ b122064

Thanks!

[dirk]

@alexcrichton: Thank you for pointing me in the right direction! 😄 That got me to take another at the parser it uses under the hood and discover the very handy .is_eof() function.

[bors]

⌛ Testing commit b122064 with merge b6ca777...

[dirk]

Looks like there's something in the build script that is causing error-index-generator to be interpreted as being passed to --cfg, which causes the compiler to error, see here.

[bors]

💔 Test failed - auto-linux-musl-64-opt

[alexcrichton]

Ah that's because that's legitimately an invalid cfg! That should probably change to error_index_generator

[dirk]

@alexcrichton: Huh, so I can see a --cfg error-index-generator on the end of the call to rustc to build src/error-index-generator/main.rs, but can't figure out where in the Makefile it's coming from.

Edit: Nevermind, found it here.

[dirk]

@alexcrichton: Turns out the fix was a simple (and probably overdue) rename! See 2766e25.

[dirk]

r? @alexcrichton

[alexcrichton]

@bors: r+ 2766e25

[bors]

⌛ Testing commit 2766e25 with merge 17d284b...

[bors]

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-debug-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-x-android-t, auto-linux-cross-opt, auto-linux-musl-64-opt, auto-mac-32-opt, auto-mac-64-nopt-t, auto-mac-64-opt, auto-mac-ios-opt, auto-win-gnu-32-nopt-t, auto-win-gnu-32-opt, auto-win-gnu-64-nopt-t, auto-win-gnu-64-opt, auto-win-msvc-32-opt, auto-win-msvc-64-opt