Rather than using this loop, use exponential backoff while waiting for
the socket path to be created.
Change-Id: I3056b1525784cd712b1d81f84876c9ca0be10dc6
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9569
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Rather than using this loop, use exponential backoff while waiting for
the socket path to be created.
Change-Id: I18706a64ce06f8916a07892dfbcd409ac5b3bff1
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9568
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
tempdir pulls in remove_dir_all 0.5.3 with
https://rustsec.org/advisories/RUSTSEC-2023-0018.html, and we use
tempfile everywhere else too, so let's just migrate to that.
Change-Id: I735ade7b65e12fc26e3d43ca95fcfa07fcc64642
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9565
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This bumps all these dependencies to their more recent versions.
We have to enable the `macros` tonic feature, and looks like we should
also enable `rt`, not only `rt-multi-thread`.
Change-Id: Icc3600848fca0bacce24f4889d088e75711594ef
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8984
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
Similar to gen_directorysvc_grpc_client, introduce a
gen_blobsvc_grpc_client function that provides a gRPC client connected
to a blobservice.
The test is update to use that client to test against, rather than the
server trait, removing the last usage of tonic_mock, so it's removed
as well.
Fixes b/243.
Change-Id: If746e8600588da247eb53a63b70fe72f139e9e77
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9564
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
vhost-user-backend doesn't support macOS yet, so the virtiofs features
will not work on macOS. This removes it as a default feature which makes
`cargo build` work out of the box on macOS.
The `virtiofs` feature is enabled for Linux when building via Nix, but
if being built by cargo directly, the feature must be enabled via a
cargo flag.
Change-Id: I2aaca9582f8e3dbcf9ee5f1b9831d614909f3799
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9555
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Autosubmit: Connor Brewster <cbrewster@hey.com>
This partially fixes b/312 and gets FUSE to work again on MacOS.
It is mostly small type changes and an update to fuse-backend-rs because
upstream currently doesn't work with MacFuse. It also sets the default
FUSE thread count on MacOS to 1 because otherwise the mount command will
hang when shutting down as only one thread gets ENODEV and all the others
just keep blocking.
Change-Id: Ifb3c4268caf296c487049c1dc4618acb32497f44
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9490
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Reviewed-by: flokli <flokli@flokli.de>
- rename the feature to `tonic-reflection` in both tvix-store and tvix-
castore.
- set it to disabled in tvix-castore by default
- enable it in tvix-store, and pull in tvix-castore with the feature
enabled.
Change-Id: Ie22833d85569502cae55812f6eeb17a9c15b9e2c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9472
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
These enable fuse and virtiofs features, which fail to build on MacOS.
tvix-cli needs neither of these to be built, so we can disable them.
Fixes b/240 (tvix-cli build on Darwin)
Change-Id: I991d947f31d0185aedd1c8a341f714f4eedd03c3
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9474
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This will bring conditional features support.
Also invoke crate2nixgenerate with the --all-features argument, so all
dependencies, including the ones for optional features are included in
the Cargo.nix file.
Change-Id: I3bbcb200c9b481f660db89efba650ea4f7418a63
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9470
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
It's in no release yet, but this is the current rev in master.
Change-Id: I9e4bc2fde26e85c4182bd66b3e566c2d957072ed
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9469
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
This adds support for listening on UNIX domain sockets via the
tokio-listener crate. The crate will automatically determine whether to
start a TCP or UNIX domain socket server based on the listen address.
Unfortunately, it's not compatible with tonic right out of the box so I
added some wrapper types to implement the necessary traits to make
things work. We should investigate upstreaming a `tonic` option to the
tokio-listener crate which implements the relevant `tonic` traits.
Example:
```
$ tvix-store daemon -l /run/tvix-store.sock
INFO tvix_store: tvix-store listening on /run/tvix-store.sock
$ tvix-store mount -l /mnt/tvix --blob-service-addr grpc+unix:///run/tvix-store.sock --directory-service-addr grpc+unix:///run/tvix-store.sock --path-info-service-addr grpc+unix:///run/tvix-store.sock
$ ls /mnt/tvix
```
Change-Id: I91c4a4b0c5a177b3b90e6c01a4e5d263130e6bdb
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9429
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
This adds a virtiofs daemon implementation which hooks into the existing
tvix-store filesystem implementation that is used for FUSE.
This allows adding the filesystem to a microvm without having to set up
FUSE inside the guest.
Change-Id: If80c36c9657f2289853e8d9a364bf4f1f7b7559c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9344
Autosubmit: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
We don't use this here, except in a test, where we don't really use
it either.
Change-Id: Ia6c45fccf663fe328942b1e1a2cd1c3a1b7f9ae5
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9376
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
This splits the pure content-addressed layers from tvix-store into a
`castore` crate, and only leaves PathInfo related things, as well as the
CLI entrypoint in the tvix-store crate.
Notable changes:
- `fixtures` and `utils` had to be moved out of the `test` cfg, so they
can be imported from tvix-store.
- Some ad-hoc fixtures in the test were moved to proper fixtures in the
same step.
- The protos are now created by a (more static) recipe in the protos/
directory.
The (now two) golang targets are commented out, as it's not possible to
update them properly in the same CL. This will be done by a followup CL
once this is merged (and whitby deployed)
Bug: https://b.tvl.fyi/issues/301
Change-Id: I8d675d4bf1fb697eb7d479747c1b1e3635718107
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9370
Reviewed-by: tazjin <tazjin@tvl.su>
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
We've decided to asyncify all of the services to reduce some of the
pains going back and for between sync<->async. The end goal will be for
all the tvix-store internals to be async and then expose a sync
interface for things like tvix eval io.
Change-Id: I97c71f8db1d05a38bd8f625df5087d565705d52d
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9369
Autosubmit: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
In prepration for adding virtiofs support, I thought it would make sense
to split out the filesystem implementation from FUSE itself.
The `fs` module holds the tvix-store filesystem implemetation and the
`fuse` module holds the code to spawn a FUSE daemon backed by multiple
threads.
Change-Id: I8c58447b8c3aa016a613068f8e7ec166554e237c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9343
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Autosubmit: Connor Brewster <cbrewster@hey.com>
This switches the FUSE implementation from fuser to fuse-backend-rs.
fuse-backend-rs is designed to work with both FUSE and virtiofs.
Virtiofs support will make it possible to plug the tvix-store into a
microvm and have `/nix/store` access without having to setup FUSE inside
the guest.
Additionally fuse-backend-rs has nice support for running multiple FUSE
threads and has some async support.
The goal of this commit is to mechanically switch over to
fuse-backend-rs with minimal changes. I did have to add some locks here
and there because fuse-backend-rs uses `&self` on all methods whereas
fuser uses `&mut self`. `&self` is required for concurrent access to the
FUSE server, so this makes sense.
We can consider switching to concurrent maps and use some other
techniques to reduce lock contention and critical section size.
Issue: https://b.tvl.fyi/issues/305
Change-Id: Icde5a58c6eef98f8984c1e04e980b756dfb76b47
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9341
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
We previously kept the trait of a BlobService sync.
This however had some annoying consequences:
- It became more and more complicated to track when we're in a context
with an async runtime in the context or not, producing bugs like
https://b.tvl.fyi/issues/304
- The sync trait shielded away async clients from async worloads,
requiring manual block_on code inside the gRPC client code, and
spawn_blocking calls in consumers of the trait, even if they were
async (like the gRPC server)
- We had to write our own custom glue code (SyncReadIntoAsyncRead)
to convert a sync io::Read into a tokio::io::AsyncRead, which already
existed in tokio internally, but upstream ia hesitant to expose.
This now makes the BlobService trait async (via the async_trait macro,
like we already do in various gRPC parts), and replaces the sync readers
and writers with their async counterparts.
Tests interacting with a BlobService now need to have an async runtime
available, the easiest way for this is to mark the test functions
with the tokio::test macro, allowing us to directly .await in the test
function.
In places where we don't have an async runtime available from context
(like tvix-cli), we can pass one down explicitly.
Now that we don't provide a sync interface anymore, the (sync) FUSE
library now holds a pointer to a tokio runtime handle, and needs to at
least have 2 threads available when talking to a blob service (which is
why some of the tests now use the multi_thread flavor).
The FUSE tests got a bit more verbose, as we couldn't use the
setup_and_mount function accepting a callback anymore. We can hopefully
move some of the test fixture setup to rstest in the future to make this
less repetitive.
Co-Authored-By: Connor Brewster <cbrewster@hey.com>
Change-Id: Ia0501b606e32c852d0108de9c9016b21c94a3c05
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9329
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
This enables the tokio `signal` feature, and registers a ctrl_c signal
handler, which will use the unmount handle to unmount in case a ctrl-c
signal is received.
This avoids having disconnected mountpoints when Ctrl-C'ing a
`tvix-store mount` invocation.
In case the filesystem is unmounted externally (via `umount /path/to/
mountpoint`), the future is waiting for the signal is never resolved and
the task is stopped.
Change-Id: I149f705a6cb50188177f2a6c6a5fcd77218e2a3f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9218
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
https://github.com/BurntSushi/walkdir/pull/170 got merged, meaning we
don't need to keep our own logic in here anymore.
Our test cases already cover this.
Change-Id: Ied3043ee651c8aafa10271c1e1ca5d460fb6c0b8
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9269
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This trait is eval-specific, there's no point in dealing with these
things in tvix-store.
This implements the EvalIO interface for a Tvix store.
The proper place for this glue code (for now) is tvix-cli, which knows
about both tvix-store and tvix-eval.
There's one annoyance with this move: The `tvix-store import` subcommand
previously also used the TvixStoreIO implementation (because it
conveniently did what we wanted).
Some of this code had to be duplicated, mostly logic to calculate the
NAR-based output path and create the PathInfo object.
Some, but potentially more of this can be extracted into helper
functions in a shared crate, and then be used from both TvixStoreIO in
tvix-cli as well as the tvix-store CLI entrypoint.
Change-Id: Ia7515e83c1b54f95baf810fbd8414c5521382d40
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9212
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: flokli <flokli@flokli.de>
Derivations can have non-unicode strings in their env values, so the
ATerm representations are not necessarily String anymore, but Vec<u8>.
Change-Id: Ic23839471eb7f68d9c3c30667c878830946b6607
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8990
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Autosubmit: flokli <flokli@flokli.de>
The change allows applications that use tvix_serde for parsing
nix-based configuration to extend the language with domain-specific
set of features.
Change-Id: Ia86612308a167c456ecf03e93fe0fbae55b876a6
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8848
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This allows constructing blob stores with a URL syntax at runtime,
by passing the --blob-service-addr arg.
We probably still want to have some builder pattern here, to allow
additional schemes to be registered.
Change-Id: Ie588ff7a7c6fb64c9474dfbd2e4bc5f168dfd778
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8742
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This brings in fuse (via the `fuser` crate), and adds pkg-config and
libfuse to the dev shell, so `cargo build` can link against it.
Change-Id: I0d11607490e27d946bdf92b0b9e45f9ab644ba74
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8664
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This switches tvix-cli over from using `NixCompatIO` to `TvixStoreIO`.
For now, we simply instantiate in-memory services instead of getting
fancy with constructors inside tvix-store, but long-term, we might want
to support some URI syntax, to make this configurable at runtime.
nixpkgs eval tests might be fine (and fast!) with a purely in-memory
backend, but other usages might involve talking to a local tvix-store
over gRPC (using the gRPC client, either unix domain socket or even
further away remote), or running tvix-store in "embedded" mode (using
another client than the gRPC client).
Change-Id: I509afd3dc5ce3f2d52b0fb7067748fab820e26ab
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8572
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Whether chunking is involved or not, is an implementation detail of each
Blobstore. Consumers of a whole blob shouldn't need to worry about that.
It currently is not visible in the gRPC interface either. It
shouldn't bleed into everything.
Let the BlobService trait provide `open_read` and `open_write` methods,
which return handles providing io::Read or io::Write, and leave the
details up to the implementation.
This means, our custom BlobReader module can go away, and all the
chunking bits in there, too.
In the future, we might still want to add more chunking-aware syncing,
but as a syncing strategy some stores can expose, not as a fundamental
protocol component.
This currently needs "SyncReadIntoAsyncRead", taken and vendored in from
https://github.com/tokio-rs/tokio/pull/5669.
It provides a AsyncRead for a sync Read, which is necessary to connect
our (sync) BlobReader interface to a GRPC server implementation.
As an alternative, we could also make the BlobReader itself async, and
let consumers of the trait (EvalIO) deal with the async-ness, but this
is less of a change for now.
In terms of vendoring, I initially tried to move our tokio crate to
these commits, but ended up in version incompatibilities, so let's
vendor it in for now.
Change-Id: I5969ebbc4c0e1ceece47981be3b9e7cfb3f59ad0
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8551
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This provides a GRPCDirectoryService struct implementing
DirectoryService, allowing a client to Directory objects from a (remote)
tvix-store.
Remote in this case is anything outside the current process, be it
another process, or an endpoint on the network.
To keep the sync interface in the `DirectoryService` trait, a handle to
some tokio runtime needs to be passed into the constructor, and the two
methods use `self.tokio_handle.spawn` to start an async function, and
`self.tokio_handle.block_on` to wait for its completion.
The client handle, called `grpc_client` itself is easy to clone, and
treats concurrent requests internally. This means, even though we keep
the `DirectoryService` trait sync, there's nothing preventing it from
being used concurrently, let's say from multiple threads.
There's still two limitations for now:
1) The trait doesn't make use of the `recursive` request, which
currently leads to a N+1 query problem. This can be fixed
by `GRPCDirectoryService` having a reference to another
`DirectoryService` acting as the local side.
I want to wait for general store composition code to pop up before
manually coding this here.
2) It's currently only possible to put() leaf directory nodes, as the
request normally requires uploading a whole closure. We might want
to add another batch function to upload a whole closure, and/or do
this batching in certain cases. This still needs some more thinking.
Change-Id: I7ffec791610b72c0960cf5307cefbb12ec946dc9
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8336
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
This stops using our own custom Hash structure, which was mostly only
used because we had to parse the JSON representation somehow.
Since cl/8217, there's a `NixHash` struct, which is better suited to
hold this data. Converting the format requires a bit of serde labor
though, but that only really matters when interacting with JSON
representations (which we mostly don't).
Change-Id: Idc5ee511e36e6726c71f66face8300a441b0bf4c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8304
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
Apparently, having multiple packages with the same path is a bad thing:
```
The bin target `tvix-store` in package `tvix-store-bin v0.1.0 (/home/flokli/tvl/tvix/store)` has the same output filename as the lib target `tvix_store` in package `tvix-store-bin v0.1.0 (/home/flokli/tvl/tvix/store)`.
Colliding filename is: /home/flokli/tvl/tvix/target/doc/tvix_store/index.html
The output filenames should be unique.
This is a known bug where multiple crates with the same name use
the same path; see <https://github.com/rust-lang/cargo/issues/6313>.
```
Change-Id: Ic785c0349070783baf5e8fd23f5fb60603a3c995
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8308
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
* We no longer need backtrace-on-stack-overflow, as we no longer
overflow the stack with the recent eval refactorings. This was weird
voodoo anyways, introduced earlier to debug some cases where stack
overflows occured.
* default features of genawaiter crate are not needed, as we don't use
their proc macros
Change-Id: I346fc5a18d7f117ee805909a8be8f535b96be76c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8263
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
This was the last piece of code using BlobWriter.
We can also use `read_all_and_chunk`, it's just requires a bit more
plumbing:
- The data coming from the client (stream) needs to be mapped (we
extract the .data field).
- The stream needs to be turned into an (async) reader
- The reader needs to be made sync, and that code using the sync reader
needs to be in a `task::spawn_blocking`.
Change-Id: I4e374e1a9f47d5a0933f59a8f5c121185a5f3e95
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8260
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This removes the use of Box::new, switching fastcdc to version 3.0.2
with https://github.com/nlfiedler/fastcdc-rs/issues/25 fixed.
Change-Id: I64f388b9e0a7f358e25a8bb7ca0e4df1d3bb01c4
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8249
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: flokli <flokli@flokli.de>
Look at the data that's written to us, and upload all chunks but the
rest in parallel, using rayon. This required moving `upload_chunk`
outside the struct, and accepting a ChunkService to use for upload
(which it was previously getting from `self.chunk_service`).
This doesn't speed up things too much for now, because things are still
mostly linear.
Change-Id: Id785b5705c3392214d2da1a5b6a182bcf5048c8d
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8195
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
This imports the contents at a given Path into the tvix store.
It doesn't register the contents at a Path in the store itself, that's up
to the PathInfoService.
Change-Id: I2c493532d65b90f199ddb7dfc90249f5c2957dee
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8159
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
This switches away from the less canonical "ronomon" version to the
implementation as described in the
[paper](https://ieeexplore.ieee.org/document/9055082) by Wen Xia, et
al., in 2020.
That version uses 64-bit hash values and tends to be faster than both
the ronomon and v2016 versions, and produces the same chunking as the
2016 version.
As per https://docs.rs/fastcdc/latest/fastcdc/#implementations-1, it's
the recommended choice.
The crate also gained support for streaming version of chunkers:
https://docs.rs/fastcdc/latest/fastcdc/#large-data, which might be
useful.
Change-Id: Ieabec3da54eb2b73c045cb54e51f7a216f63641e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8134
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
This adds a NARCalculationService trait, which will take a root node,
and provide a proto::CalculateNarResponse in return.
It also adds a NonCachingNARCalculationService implementation, that will
simply always render the NAR in memory to calculate the size and sha256,
without any caching.
Change-Id: Id1ffb18559212fa6001f70f2634bbc3dfd0aa343
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8096
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
This module contains the request/response types for generators
requesting actions from the VM.
For most of these, an async helper function is added that will be used
inside of generator functions to make use of these requests/responses
instead of constructing them directly.
Change-Id: I1e085f88adaf784a34867957a0e82532d3a83d7c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8148
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Our fork fixes a small bug (https://github.com/jneem/wu-manber/pull/1)
but it's not clear whether upstream will accept patches, so for now
lets point this directly at our fork.
Change-Id: Iccdcedae3e9a8b783241431787c952561d032694
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8031
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Switch out the string-scanning algorithm used in the reference scanner.
The construction of aho-corasick automata made up the vast majority of
runtime when evaluating nixpkgs previously. While the actual scanning
with a constructed automaton is relatively fast, we almost never scan
for the same set of strings twice and the cost is not worth it.
An algorithm that better matches our needs is the Wu-Manber multiple
string match algorithm, which works efficiently on *long* and *random*
strings of the *same length*, which describes store paths (up to their
hash component).
This switches the refscanner crate to a Rust implementation[0][1] of
this algorithm.
This has several implications:
1. This crate does not provide a way to scan streams. I'm not sure if
this is an inherent problem with the algorithm (probably not, but
it would need buffering). Either way, related functions and
tests (which were actually unused) have been removed.
2. All strings need to be of the same length. For this reason, we
truncate the known paths after their hash part (they are still
unique, of course).
3. Passing an empty set of matches, or a match that is shorter than
the length of a store path, causes the crate to panic. We safeguard
against this by completely skipping the refscanning if there are no
known paths (i.e. when evaluating the first derivation of an eval),
and by bailing out of scanning a string that is shorter than a
store path.
On the upside, this reduces overall runtime to less 1/5 of what it was
before when evaluating `pkgs.stdenv.drvPath`.
[0]: Frankly, it's a random, research-grade MIT-licensed
crate that I found on Github:
https://github.com/jneem/wu-manber
[1]: We probably want to rewrite or at least fork the above crate, and
add things like a three-byte wide scanner. Evaluating large
portions of nixpkgs can easily lead to more than 65k derivations
being scanned for.
Change-Id: I08926778e1e5d5a87fc9ac26e0437aed8bbd9eb0
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8017
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This adds a function that consumes a [proto::node::Node] pointing to the
root of a (store) path, and writes the contents in NAR serialization to
the passed [std::io::Write].
We need this in various places:
- tvix-store's calculate_nar() RPC method needs to render a NAR stream
to get the nar hash, which is necessary to give things imported in
the store a "NAR-based" store path.
- communication with (remote) Nix (via daemon protocol) needs a NAR
representation.
- Things like nar-bridge, exposing a NAR/NARInfo HTTP interface need a
NAR representation.
Change-Id: I7fb2e0bf01814a1c09094c0e35394d9d6b3e43b6
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7956
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Put this in its src/derivation.
Change-Id: Ic047ab1c2da555a833ee454e10ef60c77537b617
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7967
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Move nixbase32 and store_path into this.
This allows //tvix/cli to not pull in //tvix/store for now.
Change-Id: Id3a32867205d95794bc0d33b21d4cb3d9bafd02a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7964
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This helper function only was created because
populate_output_configuration was hard to test before cl/7939.
With that out of the way, we can pull it in.
Change-Id: I64b36c0eb34343290a8d84a03b0d29392a821fc7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7961
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This is much less code, and makes it much easier to read.
Change-Id: I9028f226105f905c2cc2cabd33907ff493e26225
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7938
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Instead of being called with `md5`, `sha1`, `sha256` or `sha512`,
`fetchurl.nix` (from corepkgs / `<nix`) can also be called with a `hash`
attribute, being an SRI hash.
In that case, `builtin.derivation` is called with `outputHashAlgo` being
an empty string, and `outputHash` being an SRI hash string.
In other cases, an SRI hash is passed as outputHash, but outputHashAlgo
is set too.
Nix does modify these values in (single, fixed) output specification it
serializes to ATerm, but keeps it unharmed in `env`.
Move this into a construct_output_hash helper function, that can be
tested better in isolation.
Change-Id: Id9d716a119664c44ea7747540399966752e20187
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7933
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This allows parsing TOML from Tvix. We can enable the eval-okay-fromTOML
testcase from nix_tests. It uses the `toml` crate, and the serde
integration it brings with it.
Change-Id: Ic6f95aacf2aeb890116629b409752deac49dd655
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7920
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Apparently our naive implementation of float formatting, which simply
used {:.5}, and trimmed trailing "0" strings not sufficient.
It wrongly trimmed numbers with zeroes but no decimal point, like
`10000` got trimmed to `1`.
Nix uses `std::to_string` on the double, which according to
https://en.cppreference.com/w/cpp/string/basic_string/to_string
is equivalent to `std::sprintf(buf, "%f", value)`.
https://en.cppreference.com/w/cpp/io/c/fprintf mentions this is treated
like this:
> Precision specifies the exact number of digits to appear after
> the decimal point character. The default precision is 6. In the
> alternative implementation decimal point character is written even if
> no digits follow it. For infinity and not-a-number conversion style
> see notes.
This doesn't seem to be the case though, and Nix uses scientific
notation in some cases.
There's a whole bunch of strategies to determine which is a more compact
notation, and which notation should be used for a given number.
https://github.com/rust-lang/rust/issues/24556 provides some pointers
into various rabbit holes for those interested.
This gist seems to be that currently a different formatting is not
exposed in rust directly, at least not for public consumption.
There is the
[lexical-core](https://github.com/Alexhuszagh/rust-lexical) crate
though, which provides a way to format floats with various strategies
and formats.
Change our implementation of `TotalDisplay` for the `Value::Float` case
to use that. We still need to do some post-processing, because Nix
always adds the sign in scientific notation (and there's no way to
configure lexical-core to do that), and lexical-core in some cases keeps
the trailing zeros.
Even with all that in place, there as a difference in `eval-okay-
fromjson.nix` (from tvix-tests), which I couldn't get to work. I updated
the fixture to a less problematic number.
With this, the testsuite passes again, and does for the upcoming CL
introducing builtins.fromTOML, and enabling the nix testsuite bits for
it, too.
Change-Id: Ie6fba5619e1d9fd7ce669a51594658b029057acc
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7922
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
This is used for content-defined chunking.
Change-Id: I10345372cecb9a643cc51ca45aa5b77d2a05198a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7889
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
These will be threaded through to eval through the new `TvixError`
variant.
Change-Id: Ia0d3f8710dcf26bb95015cd2a6a2b2911f06343f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7842
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Instead of going through Vec/BTreeMap for generating our internal
types, use the proptest strategies from imbl.
The one thing I couldn't figure out in the previous implementation is
where the ranges/sizes of generated collections came from. The
strategies in proptest use different types (Range, with an unknown
default value, and SizeRange with 0..100). I've opted to specify
0..100 directly, but we can probably make it configurable.
Change-Id: I749bc4c703fe424099240cab822b1642e5216361
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7791
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This module implements a ReferenceScanner struct which uses the
aho_corasick crate to scan string inputs for known, non-overlapping
candidates (store paths, in our case).
I experimented with several different APIs, and landed on this version
with an initial accumulator in the scanner. The scanner is
instantiated from the candidates and "fed" all the strings, then
consumed by the caller to retrieve the result.
Right now only things that look vaguely like bytestrings can be fed to
the scanner, there is no streaming support in the API yet.
Change-Id: I7782f0f0df5fc64bccd813aa14712f5525b0168c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7808
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This project was not previously covered by CI (fixed in this commit),
so we didn't catch breakage due to a renamed module.
This was noticed while rebasing a CL that has a dependency on this
crate in its Nix build.
Change-Id: Ic48570b9313e5f73e14daab50cf7ea70918c94d1
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7778
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This allows other crates to import tvix_store.
Rename the bin crate to tvix-store-bin, to avoid having multiple crates
with the same name (https://github.com/rust-lang/cargo/issues/6313)
Change-Id: I857768d6115640dbf102e79ed03e8474090df2fe
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7728
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This will make it possible fairly easily use Nix to represent
arbitrary data structures, e.g. for using Nix as a config language.
Only pure Nix (i.e. no `import` etc.) is supported for now.
Not all types, specifically no struct traversal, are implemented in
this commit.
Change-Id: I9ac91a229a0d12bf818e6e3249f3e5a691599a2c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7712
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This uses [tracing](https://github.com/tokio-rs/tracing) for logs/
tracing.
Annotate all method handlers with an instrument macro, and warn! a
message for them being unimplemented.
Co-Authored-By: Márton Boros <martonboros@gmail.com>
Change-Id: Id42a41db33782d82abfb8dc0e49a8915000e5d89
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7665
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This implements grpc.reflection.v1alpha.ServerReflection, and will make tools
like evans automatically discover available services, without having to
specify the path to the .proto files client-side.
It's behind a reflection feature flag, which is enabled by default.
Change-Id: Icbcb5eb05ceede5b9952e38a2ba72eaa6fa8a437
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7435
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This replaces the hello world example from tvix-store with an actual
gRPC endpoint, implementing all of BlobService, DirectoryService and
PathInfoService.
All RPC methods currently respond with the unimplemented gRPC status.
Co-Authored-By: Márton Boros <martonboros@gmail.com>
Change-Id: Ieba333cca44dc1e3f2ffbe676ba7a99e672b9bfb
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7664
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This implements the nix-specific base32 encoding and decoding, exposing
a subset of the API that the data-encoding crate provides.
Nix uses a custom alphabet, no padding, and encodes bytes in reverse
order. The latter one is the reason we can't just use the data-encoding
crate directly.
Three odd corner case tests ported over from go-nix failed. We opened
b/235 to further investigate.
Change-Id: I73fab6ddd67177d882e4c3f2b48761c95853d558
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7683
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This uses the `im::OrdMap` for `NixAttrs` to enable sharing of memory
between different iterations of a map.
This slightly speeds up eval, but not significantly. Future work might
include benchmarking whether using a `HashMap` and only ordering in
cases where order is actually required would help.
This switches to a fork of `im` that fixes some bugs with its OrdMap
implementation.
Change-Id: I2f6a5ff471b6d508c1e8a98b13f889f49c0d9537
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7676
Reviewed-by: sterni <sternenseemann@systemli.org>
Tested-by: BuildkiteCI
This adds a Derivation structure and allows to write it to a structure that implements std::fmt:Write.
The implementation is based on the go-nix version.
Change-Id: Ib54e1202b5c67f5d206b21bc109a751e971064cf
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7659
Reviewed-by: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This is a persistent, structurally sharing data structure which is
more efficient in some of our use-cases. I have verified the
efficiency improvement using `hyperfine` repeatedly over expressions
on nixpkgs.
Lists are not the most performance-critical structure in Nix (that
would be attribute sets), but we can already see a small (~5-10%)
improvement.
Note that there are a handful of cases where we still go via `Vec`
that need to be fixed, most notable for `builtins.sort` which can not
currently be implemented directly using `im::Vector` because of a
restrictive type bound.
Change-Id: I237cc50cbd7629a046e5a5e4601fbb40355e551d
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7670
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
lazy_static is only used in tests, and anyhow isn't used at all (yet).
This can be dropped.
Change-Id: Ic41ff3f9bb93cfa600c3485e85464f78a3976504
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7668
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
While there's currently nothing in here checking the size of the digest,
we should use something that passes the to-be-introduced validate()
function.
Change-Id: I0c515d9e3afc79292dedebce659a32485aa3d936
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7649
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
While prost-build already exposes protobuf message types as structs, we
actually need tonic-build too, to be able to get traits for all the RPC
services defined in the proto files.
Change-Id: I7f4c08454bf0d280d577975c7cdae13ccc2d933b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7320
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This type allows for temporarily compatibility with the C++ Nix store,
specifically (for now) it gives us the store directory used by Nix and
imports files the same way.
Change-Id: I4767794ef2863eba49661315c63c4e17de946d60
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7587
Reviewed-by: grfn <grfn@gws.fyi>
Tested-by: BuildkiteCI
This should make no difference in Nix builds, but allows running tests
locally again with `cargo test` for //tvix/eval.
Change-Id: I97d61840143d5c14db61d5862781bf635f9a28e7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7590
Reviewed-by: grfn <grfn@gws.fyi>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
In //tvix/eval:
* criterion bumped to 4.0, which at least depends on clap 3.x instead
of 2.x, which is less incompatible
In //tvix/cli:
* no changes required
In //tvix/nix_cli:
* some minor changes for compatibility with clap 4.0, no functionality
changes
Change-Id: If793f64b59fcaa2402d3d483ddbab4092f32df03
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7588
Reviewed-by: grfn <grfn@gws.fyi>
Tested-by: BuildkiteCI
The tvix-eval project is independent from any *uses* of the evaluator,
such as the tvix-repl.
This functionality has been split out into a separate "tvix-cli"
crate. Note that this doesn't have to mean that this CLI crate is the
"final" CLI crate for tvix, the point of this is not "getting the CLI
structure right" but rather "getting the evaluator structure right".
This reshuffling is part of restructuring the way that functionality
like store communication is injected into language evaluation.
Note that at this commit the new CLI crate is not at feature-parity.
Change-Id: Id0af03dc8e07ef09a9f882a89612ad555eca8f93
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7541
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: grfn <grfn@gws.fyi>
Tested-by: BuildkiteCI
Introduces granular dependency builds using crate2nix, bootstrapped
off the generated configuration from the newly introduced
workspace (see cl/7533).
This commit checks in the generated Cargo.nix file which can be
regenerated with a parameterless invocation of `crate2nix generate` in
`//tvix`. I tried generating this in IFD, but it turned out to be
harder than what seemed worthwhile for now.
In this setup, the various build targets for Rust projects end up
being attributes of the imported `Cargo.nix` file at the `tvix.crates`
attribute. These still lack configuration, however, which has been
fixed in the various `default.nix` files of individual projects.
Note that we (temporarily) lose the ability to build tvix-eval's
benchmarks in CI. I haven't figured out what magic incantation summons
them from the void again ...
The `eval-okay-readDir` tests from both test suites have been disabled
because they fail for unknown reasons when run in this new derivation.
Somebody will have to debug it!
Change-Id: I2014614ccb9c8951aedbd71df7966ca191a13695
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7538
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
