Commit graph

12 commits

Author SHA1 Message Date
Jiahao XU
73a794dc13
Optimization: Box fut Remote::get_redirected_final_url in GhCrateMeta::find (#600)
* Box fut `Remote::get_redirected_final_url` in `GhCrateMeta::find`
   Since the other await point in `GhCrateMeta::find` only needs
   `Arc<Self>` and `handles` to be saved, which is much smaller than the
   future returned by `Remote::get_redirected_final_url`
* Refactor: Simplify `wait_on_cancellation_signal`

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-12-12 08:42:11 +00:00
Jiahao XU
db45f2fb7f
Refactor binstalk-downloader APIs: Remove cancellation_future plus optimizations (#591)
- Refactor: Mv fn `utils::asyncify` into mod `utils`
 - Improve err msg for task failure in `utils::asyncify`
 - Make sure `asyncify` always returns the same annoymous type
   that implements `Future` if the `T` is same.
 - Rewrite `extract_bin` to avoid `block_in_place`
   support cancellation by dropping
 - Rm unused dep scopeguard from binstalk-downloader
 - Rewrite `extract_tar_based_stream` so that it is cancellable by dropping
 - Unbox `extract_future` in `async_extracter::extract_zip`
 - Refactor `Download` API: Remove `CancellationFuture` as param

   since all futures returned by `Download::and_*` does not call
   `block_in_place`, so they can be cancelled by drop instead of using this
   cumbersome hack.
 - Fix exports from mod `async_tar_visitor`
 - Make `signal::{ignore_signals, wait_on_cancellation_signal}` private
 - Rm the global variable `CANCELLED` in `wait_on_cancellation_signal`
   and rm fn `wait_on_cancellation_signal_inner`
 - Optimize `wait_on_cancellation_signal`: Avoid `tokio::select!` on `not(unix)`
 - Rm unnecessary `tokio::select!` in `wait_on_cancellation_signal` on unix
   Since `unix::wait_on_cancellation_signal_unix` already waits for ctrl + c signal.
 - Optimize `extract_bin`: Send `Bytes` to blocking thread for zero-copy
 - Optimize `extract_with_blocking_decoder`: Avoid dup monomorphization
 - Box fut of `fetch_crate_cratesio` in `PackageInfo::resolve`
 - Optimize `extract_zip_entry`: Spawn only one blocking task per fn call

   by using a mspc queue for the data to be written to the `outfile`.

   This would improve efficiency as using `tokio::fs::File` is expensive:
   It spawns a new blocking task, which needs one heap allocation and then
   pushed to a mpmc queue, and then wait for it to be done on every loop.

   This also fix a race condition where the unix permission is set before
   the whole file is written, which might be used by attackers.
 - Optimize `extract_zip`: Use one `BytesMut` for entire extraction process
   To avoid frequent allocation and deallocation.
 - Optimize `extract_zip_entry`: Inc prob of reusing alloc in `BytesMut`

   Performs the reserve before sending the buf over mpsc queue to
   increase the possibility of reusing the previous allocation.

   NOTE: `BytesMut` only reuses the previous allocation if it is the
   only one holds the reference to it, which is either on the first
   allocation or all the `Bytes` in the mpsc queue has been consumed,
   written to the file and dropped.

   Since reading from entry would have to wait for external file I/O,
   this would give the blocking thread some time to flush `Bytes`
   out.
 - Disable unused feature fs of dep tokio

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-12-12 03:15:30 +00:00
Jiahao XU
ff737730f4
Optimize use of tokio::select!: Use biased selection (#580)
as there is no need to randomize the first one to be polled.

For `cancel_on_user_sig_term` and `StreamReadable::fill_buf`, the
cancellation future should always to be polled first so that user would
feel responsive.

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-11-30 16:05:52 +13:00
Jiahao XU
50b6e62164
Minor refactor and optimization (#543)
* Avoid potential panicking in `args::parse` by using `Vec::get` instead of indexing
* Refactor: Simplify `opts::{resolve, install}` API
   Many parameters can be shared and put into `opts::Options` intead and
   that would also avoid a few `Arc<Path>`.
* Optimize `get_install_path`: Avoid cloning `install_path`
* Optimize `LazyJobserverClient`: Un`Arc` & remove `Clone` impl
   to avoid additional boxing
* Optimize `find_version`: Avoid cloning `semver::Version`
* Optimize `GhCrateMeta::launch_baseline_find_tasks`
   return `impl Iterator<Item = impl Future<Output = ...>>`
   instead of `impl Iterator<Item = AutoAbortJoinHandle<...>>`
   to avoid unnecessary spawning.
   
   Each task spawned has to be boxed and then polled by tokio runtime.
   They might also be moved.
   
   While they increase parallelism, spawning these futures does not justify
   the costs because:
    - Each `Future` only calls `remote_exists`
    - Each `remote_exists` call send requests to the same domain, which is
      likely to share the same http2 connection.
      Since the conn is shared anyway, spawning does not speedup anything
      but merely add communication overhead.
    - Plus the tokio runtime spawning cost
* Optimize `install_crates`: Destruct `Args` before any `.await` point
   to reduce size of the future
* Refactor `logging`: Replace param `arg` with `log_level` & `json_output`
   to avoid dep on `Args`
* Add dep strum & strum_macros to crates/bin
* Derive `strum_macros::EnumCount` for `Strategy`
* Optimize strategies parsing in `install_crates`
* Fix panic in `install_crates` when `Compile` is not the last strategy specified
* Optimize: Take `Vec<Self>` instead of slice in `CrateName::dedup`
* Refactor: Extract new fn `compute_resolvers`
* Refactor: Extract new fn `compute_paths_and_load_manifests`
* Refactor: Extract new fn `filter_out_installed_crates`
* Reorder `install_crates`: Only run target detection if args are valid
   and there are some crates to be installed.
* Optimize `filter_out_installed_crates`: Avoid allocation
   by returning an `Iterator`
* Fix user_agent of `remote::Client`: Let user specify it
* Refactor: Replace `UIThread` with `ui::confirm`
   which is much simpler.

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-11-19 20:00:27 +13:00
Jiahao XU
89fa5b1769
Refactor: Extract new crate binstalk-{signal, downloader} (#518)
* Refactor: Extract new crate binstalk-downloader
* Re-export `PkgFmt` from `binstalk_manifests`
* Update release-pr.yml
* Update dependabot

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-11-11 04:02:54 +00:00
Jiahao XU
fc6d9ab956
Refactor: Extract new crate binstalk-manifests (#511)
* Refactor: Extract new crate binstalk-manifests
* Fix clippy warning in mod `binstall_crates_v1`
* Rm unused deps in binstalk
* Update release-pr

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-11-08 00:50:02 +00:00
Jiahao XU
aa6012baae
Make extraction cancellable for bin and tar based formats (#481)
Extraction wasn't cancellable by `cancel_on_user_sig_term` used in `entry` since it calls `block_in_place`.

This PR adds cancellation support to it by adding a `static` variable `OnceCell` to `wait_on_cancellation_signal` so that once it returns `Ok(())`, all other calls to it after that point also returns `Ok(())` immediately.

`StreamReadable`, which is used in cancellation process, then stores a boxed future of `wait_on_cancellation_signal` and polled it in `BufReader::fill_buf`.

Note that for zip, the extraction process takes `File` instead of `StreamReadable` due to `io::Seek` requirement, so it cancelling during extraction for zip is still not possible.

This PR also optimized `extract_bin` and `extract_zip` by using `StreamReadable::copy` introduced to this PR instead of `io::copy`, which allocates an internal buffer on stack, which imposes extra copy.

It also fixed `StreamReadable::fill_buf` by ensuring that empty buffer is only returned on eof.

* Make `wait_on_cancellation_signal` pub
* Enable feature `parking_lot` of dep tokio
* Mod `wait_on_cancellation_signal`: Use `OnceCell` internally
   to archive the effect that once call to it return `Ok(())`, all calls to
   it after that also returns `Ok(())`.
* Impl `From<BinstallError>` for `io::Error`
* Impl cancellation on user signal in `StreamReadable`
* Fix err msg when cancelling during extraction in `ops::resolve`
* Optimize: Impl & use `StreamReadable::copy`
   which is same as `io::copy` but does not allocate any internal buffer
   since `StreamReadable` is buffered.
* Fix `next_stream`: Return non-empty bytes on `Ok(Some(bytes))`

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-10-13 13:31:13 +13:00
Jiahao XU
e605a99113
Impl retry for 503 and 429 status code in response (#473)
* Add new dep httpdate v1.0.2
* Enable feature time of dep tokio in binstalk
* Impl retry for `helpers::remote::Client`

Fixed #472 

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-10-10 17:54:29 +13:00
Jiahao XU
76bc030f90
Disable tcp_nodelay for reqwest::Client and add rate limiting for https requests (#458) 2022-10-07 15:51:34 +11:00
Jiahao XU
fa79e7f105
Verify that bin_files exist in resolve stage (#382)
* Refactor: Extract new fn `BinFile::check_source_exists`
* Impl new async fn `AutoAbortJoinHandle::flattened_join`
* Impl new fn `Fetcher::fetcher_name`
* Verify that `bin_files` exist in `resolve` stage
   To ensure that the installation stage won't fail because of missing
   binaries.
* Rm unused `MultiFecther`
* Simplify `Future` impl for `AutoAbortJoinHandle`
* Add new variant `BinstallError::CargoTomlMissingPackage`
* Replace `unwrap` in `resolve_inner` with proper error handling
* Make `Fetcher::new` as a regular function
   instead of an `async` function.
* Ret `Arc<dyn Fetcher>` in trait fn `Fetcher::new`
* Refactor `resolve_inner`

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-09-17 09:28:22 +00:00
Jiahao XU
934ccc257b
Support new cmdline option --roots (#389)
* Add new field `Args::roots`
* Use `env::var_os` to fetch `CARGO_INSTALL_ROOTS`
   Previously, it uses `env::var`, which might reject valid path just
   because it is not utf-8 string.
* Update manifest if `CARGO_INSTALL_ROOT` is specified
* Add new fn `install_path::get_cargo_roots_path`
* Fix updating manifest: Use `cargo_roots` instead of default path
* Rm `helpers::statics::cargo_home`

Signed-off-by: Jiahao XU <Jiahao_XU@outlook.com>
2022-09-17 20:34:36 +12:00
Félix Saparelli
e25aa50ec9
Rename lib to binstalk (#361) 2022-09-10 16:44:18 +10:00