This never did any chunking, and sled (rightfully) performs really bad
if values get too large.
We switched the default to using the objectstore backend with the local
filesystem a while ago, no need to keep this footgun around anymore.
Change-Id: I2c12672f2ea6a22e40d0cbf9161560baddd73d4a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11616
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
We already have the same code in make_grpc_path_info_service_client.
Change-Id: Ibcd60831af8a061a2c3bb2f960f52a43d06cf6fa
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11615
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This specifically allows subsequent fetch calls in the REPL to avoid
fetching the same path twice (because they share the same
PathInfoService).
Change-Id: Ieda089080174913a8c014f13d6852cac599a7e17
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11484
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Autosubmit: aspen <root@gws.fyi>
We reuse the prev_name allocation for Entry, instead of allocating and
returning a separate Vec.
We encode the `prev_name: None` case as an empty vector, since we don't
allow empty names anyway, and the sorting is equivalent.
Change-Id: I975b37ff873805f5ff099bc82128706891052247
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11607
Reviewed-by: Brian Olsen <me@griff.name>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
For small bytestrings (like NAR names), we can read into a preallocated
fixed-size buffer, instead of allocating a Vec every time.
Change-Id: Id8da9e9cea99c814361230c0ec02606b731c79a3
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11606
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
We don't actually build up names in place here, so we don't need a
capacity field. Saves 8 bytes.
Change-Id: Icb01b45561e28fd525b726612f56d4640bc834c7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11604
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
If our underlying reader supports AsyncBufRead, then we can too.
Change-Id: If4b948c983400ca591c1c475bbcf7dc00d562040
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11545
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Even if we have an aligned (or zero) size (and thus have no padding),
we may still have a non-padding trailer.
Since an exact read of only the user data bytes must always read the
trailer bytes as well, this has to happen for aligned sizes too.
For zero-sized trailers, we solve this by reading immediately, since no
read is required of the user at all.
user_len thus becomes a NonZeroU64, and the definition of "body data"
is adjusted to always exclude the final block.
Change-Id: I2c6bb51270fb22ad72a65076a3468673ab56ef68
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11597
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
We're using this in the NAR reader now.
Change-Id: I28f17b1ccedd62ffcaf2fa32b517f16bcd036d94
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11603
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
We specifically structured the code this way to allow using
`this.filled` as-is, so we should use it.
Change-Id: I7e11bddceb4d5f37b1dd4c453a9d53b85fc1f6c8
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11602
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Rust 1.77 supports async recursion as long as there is some form of
indirection (ie. `Box::pin`). This removes the need to use the
async-recursion crate.
Change-Id: Ic9613ab7f32016f0103032a861edff92e2fb8b41
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11596
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Rename read_nar to ingest_nar, and have it use the async nar reader
version, and the ingest_entries machinery.
This means we can now drop all code dealing with manually assembling
castore nodes.
Update our consumer, NixHTTPPathInfoService to use the new API.
As we now accept an AsyncRead, we don't need to do any blocking here
anymore, and can use the same async-compression crate as in the fetching
logic (and support some more compression formats out of the box).
Change-Id: I8646d20bd8603f8da47b5c84bc9e4ac236eb7f1a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11580
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Once we break out with the root node, there may be no more elements in
the stream.
Change-Id: I6f5fc5662095aa2b2a56bcad506d25520d9ad00c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11592
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
We got away with not properly dealing with this for the archive case,
where everything is contained inside a toplevel dir, but NARs can encode
a single file/symlink.
Properly break if the IngestionEntry path has the ROOT as parent, and
only create filling directories in the other case.
Change-Id: Ib378d0d1040de7c3fe310912a0b0488c55afee83
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11590
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
There's no need for this to be a &PathBuf.
Change-Id: I2d4126d57cfd8ddaad5dd327943b70b83d45c749
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11589
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
This is a first cut at the async NAR reader, with some rough edges.
Poisoning is left unimplemented for now, pending future work.
Change-Id: Ifaafe0581a5e0e165a13357b909fb441f7bd8bab
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11524
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This allows using BytesReader with a custom tag, eg the closing parens
for the NAR reader.
No public constructor is provided for custom-tagged readers, since this
feature isn't public API.
Change-Id: I82e73d064edc4b6783ead1d6fe46a5b35f45c844
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11543
Reviewed-by: Brian Olsen <me@griff.name>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
We implement DirectoryService for Arc<DirectoryService> and
Box<DirectoryService>, this is sufficient.
Change-Id: I0a5a81cbc4782764406b5bca57f908ace6090737
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11586
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Write this a bit more compact, by using map_err(|e| …) and ?.
Ideally we'd get rid of the error mapping entirely, by using proper
error types, but that's left for a followup.
Change-Id: I68dc72b162ac89c5ff82d8c2bc26e1c808a0affd
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11584
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
The emulator and bigtable client are quite big. Remove them from the
default //tvix:shell.
Put the tests behind a `integration` feature flag, and add a variant
with that enabled to CI, and drop the bigtable tools from //tvix:shell.
Change-Id: Ie042097a0d6fc26542faa96c139b77298ccb160a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11582
Reviewed-by: edef <edef@edef.eu>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Since a recent nixpkgs bump bringing a version of cargo with
https://github.com/rust-lang/cargo/pull/12914,
crate2nix creates a crate-hashes.json with all crate hashes from
Cargo.lock (and downloads a lot of stuff while producing it).
https://github.com/nix-community/crate2nix/pull/341 prevents this from
happening, but our hardcoded crate2nix pin prevented us from getting the
fix included in 0.14.0, which did land in nixpkgs.
Replace the pin with a simply override, carrying our only leftover patch
on top of it, and link to that PR.
Change-Id: I9503898e15d61fa6a2b1589d141bec1b4ed3d616
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11581
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
This documents which Nix version uses which daemon protocol version so that
any implementor of the protocol can make informed choices about how far back to
support legacy.
Change-Id: If2fc11702f7f731816d51928efc9ea9d939e1c00
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11579
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This switches from using std::path::Path to using castore paths.
We can drop some error handling in descend_to, as absolute (or redundant)
paths are not representable.
We however now need to convert from a std::path::Path to our
representation, and decide to accept .. canonicalization, as paths in
EvalIO might contain this. Dealing .. to hop into another store path, if
we encounter this, should be dealt with in a previous step.
Change-Id: I5e94693808420c5d56587c68731252b54755bf93
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11575
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
This allows converting from std::path::Path to castore PathBufs.
A flag is present to control .. canonicalization, and the usual caveats
about platform-specific differences apply.
Currently only added for unix, we'll carefully consider other platforms
on a case-by-case basis.
Change-Id: If289a92f75a2e5c3eec132b6a91a28d225fc1989
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11577
Reviewed-by: edef <edef@edef.eu>
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
There's no need for us to come up with our own names for this.
Also update the comments/docstrings a bit, and inline the intermediate
hash_derivation_modulo calculation.
Change-Id: I09dab8ffe1ebfb6601841e98119eee4ff25d8f39
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11578
Reviewed-by: edef <edef@edef.eu>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This allows using both Path and PathBuf in a function argument taking
`impl AsRef<Path>`.
Change-Id: Ibd3ba6fac538069d2fe729d1ef399fdef301668f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11574
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
These are fallible methods, so they should be named accordingly.
Change-Id: I6dc271c42989dd6500173488190f65381835d6fe
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11572
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
The empty path (Path::ROOT) is explicitly a valid path, and "foo" is
simply a child of "". The root itself is the only path without a parent.
Change-Id: Iff00dc8aed89eaf98702b664c0df658bd5a1d88a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11569
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Use the shared code for validating node names, since that is what path
components represent.
Change-Id: I12109c1306b224718faa66cf1f2874c78c1436a7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11566
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Since PathBuf doesn't have inherent methods anymore, these just forward
to Path itself.
Change-Id: I30f44adc9994337c367bad985ada0e8fcb98dd6a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11570
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Nix 2.4 accidentally permitted this behaviour, but the revert came
too late to beat Hyrum's law. It is now considered permissible.
Link: https://github.com/NixOS/nix/pull/9867
Change-Id: Ie97777af6765fe1c12c8aa593afe1c9b69125775
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11553
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
We implement Debug explicitly, so that we don't just see raw integers.
Change-Id: I11213094728f3e0c674562ee71c092a950041632
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11565
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This contains Path and PathBuf, representing platform-independent paths
representable by the castore model.
These are always relative, and platform-independent, which distinguishes
them from the ones provided in the standard library.
A subsequent CL will move IngestionEntry (and more) to use them.
Change-Id: Ib85857f4159ebc2f3c00192c95d4e5b54ffd4fcf
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11558
Tested-by: BuildkiteCI
Reviewed-by: edef <edef@edef.eu>
The sled backend doesn't perform very well with blobs in there,
especially as it's not doing any chunking.
Switch to the `objectstore+file://` instead, which does do CDC.
Change-Id: Ic0d8836c6fc811b80c7202e3ee7f44a05a4f8dfa
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11554
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Have ingest_entries return an Error type with only three kinds:
- Error while uploading a specific Directory
- Error while finalizing the directory upload
- Error from the producer
Move all ingestion method-specific errors to the individual
implementations.
Change-Id: I2a015cb7ebc96d084cbe2b809f40d1b53a15daf3
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11557
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
We shouldn't try to represent non-representable things in the ingestion
entries (only to throw an error).
It's cleaner to throw the error directly in the part producing the
stream.
Change-Id: I6b6f6d8c2f677425210142a39f1829ddeefec812
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11556
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: firefly <firefly@firefly.nu>
This is only useful for when we have access to a filesystem, so it
shouldn't be in the root.
Change-Id: I9923aaed1aef9d3a1e8fad41f58821d51c2eb34b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11555
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: firefly <firefly@firefly.nu>
Tested-by: BuildkiteCI
These can be arbitrary bytes in theory. Some of our libraries might
be more strict, or inconsistent w.r.t. their representation of path
separators.
Change-Id: I7981b74fc7d3dd79f5589cf2ef52ced7b71dd003
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11551
Tested-by: BuildkiteCI
Reviewed-by: edef <edef@edef.eu>