xdBronch

`std.mem.readInt` doesnt branch at runtime because its an inline function which has the same semantics proposed here but for all parameters, thats why this was unaccepted (at least i can...

the output of `--help` relies on being able to run your build file since it displays user options and steps

iiuc `await` is only used when the coroutine is completed but you still have suspension points. adding 2 calls to `libcoro.xresume(frame);` before the call to `xawait` makes the test case...

shouldnt this be a compile error from the `type_info.Fn.return_type.?`? the return type is generic so iiuc it should be null

@reader-lamp the benchmarks you keep doing are not representative and really not useful, the point of incremental is making *rebuilds* faster and you didn't even use `--watch` which is currently...

> they have repeatedly shown Zig being orders of magnitude slower than other options they show zig being slower in a comparison that is not representative. Go and Zig are...

there are some more breaking changes on master, https://github.com/ziglang/zig/pull/21225 ~~and possibly https://github.com/ziglang/zig/pull/21231 eventually (havent checked if this one affects zls)~~(it doesnt). could maybe do them all here so there arent...

this works properly now

imo even though the usage is wrong it seems odd to purposefully make this not work, the `try` is wrong not the `.foo`. i see it as being similar to...

this could be pretty easily fixed by using something like [`nodesOverlappingIndex`](https://github.com/zigtools/zls/blob/73c5a10ab2e167006709e2ecf242b336ca2f1f4a/src/ast.zig#L1557) [here](https://github.com/zigtools/zls/blob/73c5a10ab2e167006709e2ecf242b336ca2f1f4a/src/features/references.zig#L161) but i can only imagine how potentially expensive that might end up on large files, though it might...