Commit Graph

4 Commits

Author SHA1 Message Date
Jakub Konka
bd5e71b038 [wasi-common]: add armv7 support to wasi-common (#1269)
* Add armv7 support to wasi-common

This commit enables `target_pointer_width = 32` compatibility for
`wasi-common` (and by transitivity, any crate found inside, e.g., `yanix`).
I've also added a simplistic (bare minimum) check to our CI to ensure
that `wasi-common` cross-compiles to `armv7-unknown-gnueabihf` fine.
While here, I've done the same for `wasm32-unknown-emscripten`.

* Clean arch-specific impls + reuse libc consts

* Make SeekLoc::from_raw platform independent

* Collapse CI cc jobs into one
2020-03-10 19:18:59 +01:00
Jakub Konka
061390ee1b [wasi-common]: move filetime module to yanix (#1255)
* Move filetime module to yanix

I've noticed that we could replace every occurrence of `crate::Result`
in `filetime` mods with `io::Result`, so I thought why not move it
to `yanix` and get rid off a lot of unnecessary code duplication
within `wasi-common`. Now, ideally I'd have our `filetime` modifications
backported to Alex's [`filetime`] crate, but one step at a time
(apologies Alex, I was meant to backport this ages ago, just didn't
find the time yet... :-().

Anyway, this commit does just that; i.e., moves the `filetime` modules
into `yanix` which seems a better fit for this type of code.

[`filetime`]: https://github.com/alexcrichton/filetime

There is one caveat here. On Emscripten, converting between `filetime::Filetime`
and `libc::timespec` appears to be lossy, at least as far as the
types are concerned. Now, `filetime::Filetime`'s seconds field is
`i64` while nanoseconds field is `u32`, while Emscripten's
`libc::timespec` requires both to be `i32` width. This might actually
not be a problem since I don't think it's possible to fill `filetime::Filetime`
struct with values of width wider than `i32` since Emscripten is 32bit
but just to be on the safe side, we do a `TryInto` conversion, log
the error (if any), and return `libc::EOVERFLOW`.

* Run cargo fmt

* Use i64::from instead of as cast
2020-03-09 16:07:09 +01:00
Jakub Konka
42fae4e3b8 [wasi-common]: yanix now returns io::Error directly (#1242)
* Yanix now returns io::Error

This commit may seem somewhat controversial at first, but hear me
out first. Currently, Yanix would return a custom error that's a
wrapper around three other error types returned by various entities
inside Rust's `libstd`. In particular, Yanix's error type would wrap
`io::Error`, `num::TryFromIntError` and `ffi::NulError`. It turns
out that there is a natural conversion between the first and the last
and provided by the standard library, i.e., `From<ffi::NulError> for io::Error`
is provided. So at the surface it may seem that only the first two
wrapped error types are worth keeping.

Digging a little bit deeper into `libstd`, `num::TryFromIntError`
is essentially speaking only a marker that the integral conversion
went wrong. The struct implementing this error stores a unit type,
and nothing more. It therefore seems like a waste to wrap this
particular error when we could unify everything under `io::Error`.
And so, whenever we perform an int conversion, I suggest we simply
remap the error to `io::Error::from_raw_os_error(libc::EOVERFLOW)`
since this carries a comparable amount of information.

As a result of completely discarding `yanix::Error` custom error type,
we are invariably simplifying `yanix` itself, but also allowing
`wasi-common` to simplify in several places as well.

* Adapt wasi-common to changes in yanix

* Add Cargo.lock

* Unwrap try_into's where possible

* Remove unnecessary type annotation
2020-03-06 14:20:54 -08:00
Jakub Konka
95c2addf15 Compile wasi-common to Emscripten (#688)
* Compile wasi-common to Emscripten

This commit enables cross-compiling of `wasi-common` to Emscripten. To achieve
this, this commit does quite a bit reshuffling in the existing codebase. Namely,
* rename `linux` modules in `wasi-common` and `yanix` to `linux_like` -- this is
  needed so that we can separate out logic specific to Linux and Emscripten out
* tweak `dir` module in `yanix` to support Emscripten -- in particular, the main
  change involves `SeekLoc::from_raw` which has to be now host-specific, and is now
  fallible
* tweak `filetime` so that in Emscripten we never check for existence of `utimensat`
  at runtime since we are guaranteed for it to exist by design
* since `utimes` and `futimes` are not present in Emscripten, move them into a separate
  module, `utimesat`, and tag it cfg-non-emscripten only
* finally, `to_timespec` is now fallible since on Emscripten we have to cast number of
  seconds, `FileTime::seconds` from `i64` to `libc::c_long` which resolves to `i32`
  unlike on other nixes

* Fix macos build

* Verify wasi-common compiles to Emscripten

This commit adds `emscripten` job to Github Actions which installs
`wasm32-unknown-emscripten` target, and builds `wasi-common` crate.

* Use #[path] to cherry-pick mods for Emscripten

This commit effectively reverses the reorg introduced in 145f4a5
in that it ditches `linux_like` mod for separate mods `linux` and
`emscripten` which are now on the same crate level, and instead,
pulls in common bits from `linux` using the `#[path = ..]` proc
macro.
2019-12-11 16:25:13 -08:00