This change includes only the basic nix handshake protocol handling and
sets up a client session. The only supported operation at this point is
SetOptions.
Additional operations will be implemented in subsequent cls.
Change-Id: I3eccd9e0ceb270c3865929543c702f1491768852
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12743
Autosubmit: Vladimir Kryachko <v.kryachko@gmail.com>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Reviewed-by: edef <edef@edef.eu>
Reviewed-by: Brian Olsen <me@griff.name>
tvix-castore already supports composition without any additional feature
flags, the only thing that can be explicitly enabled is referring to
other stores via an anonymous url. Rename that feature flag to
"xp-composition-url-refs".
tvix-store effectively only controls the CLI surface, so rename this to
"xp-composition-cli".
The "store" in the feature name was dropped, as it's already apparent
from being in the tvix-[ca]store crate.
Change-Id: I1175dc6280cbba4cbcdfb7fd4b35fce713b45fc4
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12747
Tested-by: BuildkiteCI
Reviewed-by: yuka <yuka@yuka.dev>
This change implements the serialization part that is needed to
implement the nix daemon protocol. Previously was add deserialization
and derivers for that and this then adds the other part of that equation
so that you can write types that can then be read using deserialization.
Change-Id: I2917de634980a93822a4f5a8ad38897b9ce16d89
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12729
Autosubmit: Brian Olsen <me@griff.name>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
The goal is to create a drop-in replacement nix-daemon that nix-cpp can
use as a `daemon` store.
Change-Id: Ie092047dcc6a24a3b8d8d1b808f3e6fd2c493bf2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12711
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This is only using nix-compat through tvix-glue.
Change-Id: Ib6391987954a28cc05a2c27c1c21585185d82d1a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12681
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Ilan Joselevich <personal@ilanjoselevich.com>
Tested-by: BuildkiteCI
This is only using tvix-castore through tvix-store.
Change-Id: I342ee1e18f8c3946b712cfb446520b732091d58f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12680
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Ilan Joselevich <personal@ilanjoselevich.com>
This implements BS, DS, PS for Box'ed or Arc'ed variants of it with less
code, and less potential to accidentially forget to proxy default trait
methods for blanked impls, as fixed in cl/12658.
Change-Id: If2cdbb563a73792038ebe7bff45d6f880214855b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12661
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: edef <edef@edef.eu>
This is now supported in the standard library via std::sync::LazyLock,
but requires some manual shuffling around of code.
Change-Id: Ifca792f4d2dbc36b703de4a4dfa406015ab86da7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12614
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: flokli <flokli@flokli.de>
Reviewed-by: edef <edef@edef.eu>
Tested-by: BuildkiteCI
This is now supported in the standard library via std::sync::LazyLock,
but requires some manual shuffling around of code.
Change-Id: I14bee4068dc73c948321481b5a4e1fc922a89a27
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12611
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
This is now supported in the standard library via std::sync::LazyLock,
but requires some manual shuffling around of code.
Change-Id: Ie2af74beda9fcf8aa19fca7d844bcbe732f05bf8
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12610
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This is now supported in the standard library via std::sync::LazyLock, but
requires some manual shuffling around of code.
Change-Id: Ibb3be8458b8a8912ea04c9360d64c5cf914254d4
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12609
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This is now supported in the standard library via std::sync::LazyLock, but
requires some manual shuffling around of code.
I found at least one dead variable along the way, which I deleted.
Change-Id: I8600c87c49078fb5ff72671994c77b919259e67b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12608
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
Equivalent logic is now in the standard library, and this dependency is no
longer needed for eval.
Change-Id: Iaa4410d89fdaa5b84cbd9e6bc6ae479c659d92f2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12602
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: edef <edef@edef.eu>
Tested-by: BuildkiteCI
Our oci-spec was a bit oudated and there were some renamings in one of
the release, which made building tvix-build fail if it's a dependency.
I encountered this issue while working on tvix-eval-jobs.
Change-Id: I6d982965176b83170a07445e351d3f5e5679ed2e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12586
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
Tested-by: BuildkiteCI
This allows specifying an url in place of a named reference to another
composition entry, if the castore crate has been compiled with the
xp-store-composition feature.
Example: `--directory-service-addr cache://?near=memory://&far=memory://`
This would be equivalent to the instantiation via toml file:
```toml
[memory1]
type = "memory"
[memory2]
type = "memory"
[default]
type = "cache"
near = "memory1"
far = "memory2"
```
Note that each anonymous url causes a distinct instance to be created.
Change-Id: Iee5a07a94b063b5e767c704d9cad0114fa843164
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12146
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
We wanted to enable this earlier but the log level of many of the
messages in redb were too high, they've now been downgraded, so we can
enable logs from redb.
Context on the fix and release:
- https://github.com/cberner/redb/pull/828
- https://github.com/cberner/redb/releases/tag/v2.1.2
Change-Id: I8635e8a0bcb01a7d0b580387ac9134ccdd0205f3
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12568
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Over the past couple of months we've been using redb instead of sled as
the default filesystem-based database in PS and DS. I am confident that
we can get rid of sled completely now, and just keep redb.
Change-Id: I11fa1e4453e280253855f8eade990b37eb6965ae
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12567
Reviewed-by: yuka <yuka@yuka.dev>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
This is just patchset 10 of CL10855, before the color_eyre changes,
rebased to the tvix_castore api.
Change-Id: If4b42412ff8568058908cda971ad7d6f2d9f9b7b
---
This provides a build service invoking runc. It can be used by using the
`oci://$path_to_some_tempdir` builder URL for now.
For now, it can be tested as such:
```
BUILD_SERVICE_ADDR=oci://$PWD/bundles target/debug/tvix
let pkgs = (import <nixpkgs> {}); in builtins.readDir pkgs.perl
```
readDir is to actually trigger IO into the store path (which triggers
the builds).
For now it fails due to missing reference scanning (see followup CLs).
Change-Id: I09b40e410114ce69966a41a0e3c33281b859e443
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12526
Autosubmit: yuka <yuka@yuka.dev>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
With an implementation of AsyncRead + AsyncSeek, axum-range can answer
range requests.
We only use it if a range has been requested, as it uses more memory
than the linear variant.
Change-Id: I0072b0a09b328f3e932f14567a2caa3a49abcbf7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12509
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Reviewed-by: yuka <yuka@yuka.dev>
This release includes support for overlay filesystems
Change-Id: I946cbf346df045209afaea2d720bb57fc2f2659f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12538
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: yuka <yuka@yuka.dev>
Tested-by: BuildkiteCI
This is required to add the optional refscanner parameter to the
ingest functions.
Change-Id: Ib40a7287cf857eb55e31e0df309a79474fefb518
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12527
Autosubmit: yuka <yuka@yuka.dev>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
- Add a new PathInfoService implementation that wraps transparently
around another except that it dynamically signs all the incoming
path-infos with the provided signer.
- Add a ServiceBuilder for this PathInfoService that provides a
SigningPathInfoService with a keyfile signer
Change-Id: I845ddfdf01d14c503c796b2b80c720dab98be091
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12032
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: sinavir <tvix@sinavir.fr>
Tested-by: BuildkiteCI
thiserror is much more easier to maintain than manually implementing Error and Display.
Change-Id: Ibf13e2d8a96fba69c8acb362b7515274a593dfd6
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12452
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
Tested-by: BuildkiteCI
In https://cl.tvl.fyi/c/depot/+/12389 I accidentally deleted some
comments. Also fixed some useless no-default-features with futures crate, making it match the previous feature combination.
Change-Id: I72bb2cfb88719ff2f8812d90193de2bd49149cce
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12395
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
From now on we will add the dependencies and their version in the root
Cargo.toml and in order to enable the dependency for a workspace member
we set `workspace = true` in the member's Cargo.toml.
Change-Id: I9738c1cf99810b7ace87ca712c3ea965ba846e25
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12389
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This updates all the dependencies and their "minimum" versions in
Cargo.{lock,toml} to the latest compatible version using `cargo-edit`'s
`cargo upgrade` command that will eventually be merged into `cargo
update`.
Change-Id: Iccb2aa4a1c84a0465222244a0bd0cafe2a82e781
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12388
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
Tested-by: BuildkiteCI
This bumps bigtable_rs to
https://github.com/liufuyang/bigtable_rs/pull/86, allowing us to drop
our second set of prost/tonic/http/axum crates.
Change-Id: I70f9150289c3e8611ebe8a7d99490e3dfd085a6e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12384
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
The external feature flag was there because I couldn't find a way to
refer to crate and nix-compat with the same name so that the generated
code could be the same.
In essence `use nix_compat::nix_daemon:🇩🇪:NixDeserialize` is an error
when used inside nix_compat crate.
So my best fix was the external feature flag until I found the solution
used here which also removes the flag completely.
Change-Id: Ia3e89c6c350c3fb22ca87f974a39c21542aae152
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12376
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: Brian Olsen <me@griff.name>
tonic-reflection 0.12.x moved from the v1alpha to v1 of the reflection
protocol.
However, most clients, like Postman, Kreya and evans don't support that
one yet.
Bump tonic-reflection to 0.12.2, which re-introduces v1alpha support
alongside the v1 version of it, registering both services.
This fixes the example documented in tvix/store/README.md, it was
previously failing as evans couldn't find the v1alpha reflection
service.
See https://github.com/hyperium/tonic/pull/1888 for details.
Change-Id: I55438877317f82dc39face13afeb9594cda07a4e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12353
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Ilan Joselevich <personal@ilanjoselevich.com>
This adds a nix-compat-derive derive crate that implements a deriver
for NixDeserialize implementations. This is to reduce the amount of
code needed to implement deserialization for all the types used by
the Nix daemon protocol.
Change-Id: I484724b550e8a1d5e9adad9555d9dc1374ae95c2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12022
Autosubmit: Brian Olsen <me@griff.name>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Add a trait for deserializing a type from a daemon worker connection.
This adds the NixDeserialize trait which is kind of like the serde
Deserialize trait in that individual types are meant to implement it
and it can potentially be derived in the future.
The NixDeserialize trait takes something that implements NixRead as
input so that you can among other things mock the reader.
Change-Id: Ibb59e3562dfc822652f7d18039f00a1c0d422997
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11990
Autosubmit: Brian Olsen <me@griff.name>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This moves the URL component parsing code we had in nar-bridge to
nix-compat.
We change the function signature to return an Option, not a
Result<_, StatusCode>.
This allows returning more appropriate error codes, as we can
ok_or(…) at the callsite, which we now do: on an upload to an
invalid path, we now return "unauthorized", while on a GET/HEAD, we
return "not found".
This also adds support to parse compression suffixes. While not
supported in nar-bridge, other users of nix-compat might very well want
to parse these paths.
Also fix the error message when parsing NAR urls, it mentioned 32, not
52, which is a copypasta error from the narinfo URL parsing code.
Change-Id: Id1be9a8044814b54ce68b125c52dfe933c9c4f74
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12260
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Requiring `name` to be a `&str` means it'll get annoying to pass around
`Signature`, but being able to pass them around in an owned fashion is
kinda a requirement for a stronger typed `PathInfo` struct, where we
want to have full ownership.
Rework the `Signature` struct to become generic over the type of the
`name` field. This means, it becomes possible to have owned versions
of it.
We don't want to impose `String` or `SmolStr` for example, but want to
leave it up to the nix-compat user to decide.
Provide a type alias for the existing `&str` variant (`SignatureRef`),
and use it where we previously used the non-generic `Signature` one.
Add some tests to ensure it's possible to *use* `Signature` with both
`String` and `SmolStr` (but only pull in `smol_str` as dev dependency
for the tests).
Also, add some more docstrings, these were a bit sparse.
Change-Id: I3f75691498c6bda9cd072d2d9dac83c4f6c57287
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12253
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Removes imbl::OrdMap in favour of an Rc over the standard library's BTreeMap,
which allows us to drop the imbl dependency completely.
In my local tests this is actually slightly faster for `hello` and `firefox`.
Change-Id: Ic9597ead4e98bf9530f290c6a94a3c5c3efd0acc
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12201
Reviewed-by: aspen <root@gws.fyi>
Tested-by: BuildkiteCI
This replaces the OpCode enum with a new Op enum which is guaranteed to fit in a
single byte. Instead of carrying enum variants with data, every variant that has
runtime data encodes it into the `Vec<u8>` that a `Chunk` now carries.
This has several advantages:
* Less stack space is required at runtime, and fewer allocations are required
while compiling.
* The OpCode doesn't need to carry "weird" special-cased data variants anymore.
* It is faster (albeit, not by much). On my laptop, results consistently look
approximately like this:
Benchmark 1: ./before -E '(import <nixpkgs> {}).firefox.outPath' --log-level ERROR --no-warnings
Time (mean ± σ): 8.224 s ± 0.272 s [User: 7.149 s, System: 0.688 s]
Range (min … max): 7.759 s … 8.583 s 10 runs
Benchmark 2: ./after -E '(import <nixpkgs> {}).firefox.outPath' --log-level ERROR --no-warnings
Time (mean ± σ): 8.000 s ± 0.198 s [User: 7.036 s, System: 0.633 s]
Range (min … max): 7.718 s … 8.334 s 10 runs
See notes below for why the performance impact might be less than expected.
* It is faster while at the same time dropping some optimisations we previously
performed.
This has several disadvantages:
* The code is closer to how one would write it in C or Go.
* Bit shifting!
* There is (for now) slightly more code than before.
On performance I have the following thoughts at the moment:
In order to prepare for adding GC, there's a couple of places in Tvix where I'd
like to fence off certain kinds of complexity (such as mutating bytecode, which,
for various reaons, also has to be part of data that is subject to GC). With
this change, we can drop optimisations like retroactively modifying existing
bytecode and *still* achieve better performance than before.
I believe that this is currently worth it to pave the way for changes that are
more significant for performance.
In general this also opens other avenues of optimisation: For example, we can
profile which argument sizes actually exist and remove the copy overhead of
varint decoding (which does show up in profiles) by using more adequately sized
types for, e.g., constant indices.
Known regressions:
* Op::Constant is no longer printing its values in disassembly (this can be
fixed, I just didn't get around to it, will do separately).
Change-Id: Id9b3a4254623a45de03069dbdb70b8349e976743
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12191
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This splits the existing ReferenceScanner into a ReferenceScanner and
ReferencePattern as well as adds an AsyncRead implementation that can
do a scan while you read from it.
The reason to split the scanner in two is that generating the pattern
is expensive and when ingesting build results with multiple outputs you
want to do several independant scans that look for the same pattern.
The reader is for scanning files without having to load the entire file
into memory.
Change-Id: I993f5a32308c12d9035840f8e04fe82e8dc1d962
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12052
Autosubmit: Brian Olsen <me@griff.name>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Rather than storing the leaked allocation for the string as the key in
the interner, store the hash (using NoHashHashBuilder). I thought this
would improve performance, but it doesn't:
hello outpath time: [736.85 ms 748.42 ms 760.42 ms]
change: [-2.0754% +0.4798% +2.7096%] (p = 0.72 > 0.05)
No change in performance detected.
but it at least doesn't *hurt* performance, and it *does* avoid an
`unsafe`, so it's probably net good.
Change-Id: Ie413955bdb6f04b1f468f511e5ebce56e329fa37
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12049
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: aspen <root@gws.fyi>
The problem with using runCommand and recreating the src directory with
lndir is that it changes the file types of individual files, they will
now be a symlink instead of a regular file. If you have a crate that tests
that a file is of regular type then it will fail inside the crate2nix
derivation.
Also regenerate Cargo.nix for //tvix as it will be needed in the next
commit.
Change-Id: I9275602cc17a428f9fdf0e55daf12cd673bbc030
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12131
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Use the faster and newer MiMalloc memory allocator for all endpoints in
the workspace.
Change-Id: Ic60237284ed168e46ec6e8f28e2710bae4385c6f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12149
Tested-by: BuildkiteCI
Reviewed-by: aspen <root@gws.fyi>
Per https://nnethercote.github.io/perf-book/hashing.html, we have
basically no reason to use the default hasher over a faster,
non-DoS-resistant hasher. This gives a nice perf boost basically for
free:
hello outpath time: [704.76 ms 714.91 ms 725.63 ms]
change: [-7.2391% -6.1018% -4.9189%] (p = 0.00 < 0.05)
Performance has improved.
Change-Id: If5587f444ed3af69f8af4eead6af3ea303b4ae68
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12046
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Reviewed-by: Ilan Joselevich <personal@ilanjoselevich.com>
Autosubmit: aspen <root@gws.fyi>
Default to always leaking strings, and copying strings by copying
pointers rather than cloning the underlying allocation. This (somewhat
bafflingly) doesn't seem to affect any benchmarks, but paves the way for
some tricks around string allocation that do.
Unfortunately, we can't do this (yet?) for contextful strings, for
reasons I don't currently understand but which I will address later,
when I address contextful strings more holistically.
I've left a flag in here to disable this, both to test the cloning logic
for strings for when/if we decide to bring this back, and to allow
people who care more about memory usage than perf to disable leaking.
Change-Id: Iec44bcbfe9b3d20389d2450b9a551792a79b9b26
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12045
Autosubmit: aspen <root@gws.fyi>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This provides a PathInfoService implementation using redb
(https://github.com/cberner/redb) as the underlying storage engine.
Both an in-memory variant, as well as a filesystem one is provided,
similar how it's done with the sled implementation.
Supersedes: https://cl.tvl.fyi/c/depot/+/11692
Change-Id: I744619c51bf2efd0fb63659b12a27cbe0b2fd6fc
Signed-off-by: Ilan Joselevich <personal@ilanjoselevich.com>
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11995
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
