This reverts commit f5e291cf83.
The offsets are relative to the start of the file, and as long as we
don't have BlobReaders implement seek, this will be very annoying to
deal with.
Change-Id: I05968f7c5c0ec0000597da90f451d6bb650c3e13
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8882
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
buf contains everything written so far, whereas b is the slice passed in
the current write() call. If we copy from &buf, we end up with the wrong
hash, because we keep writing the wrong data to the hash function.
Change-Id: I768d4645934a6a7d75b9c8eeba35f8f3be5edd26
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8880
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
This moves from stateless I/O to actually dealing with file handles,
allowing the filesystem to keep reusing existing blobreaders, instead of
opening a new reader on every read() call.
Change-Id: I3fc35c071e4aee1021c8bbd58749d082b0abd188
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8834
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: flokli <flokli@flokli.de>
* //3p/sources: switch stable channel from 21.11 (!) to 23.05
* //users: adapt to emacsUnstable to emacs-unstable rename
* //users/grfn: use default Linux kernel version everywhere,
as 5.15 has broken in this version of nixos-unstable.
* //3p/cgit: adapt to git 2.41.0
The committed changes are the same as the [patch1] I've submitted
to cgit-pink which is in turn based on Christian Hesse's [patch2].
patch1: https://causal.agency/list/thread/20230624144033.802270-1-sternenseemann%40systemli.org.html#20230624144033.802270-2-sternenseemann@systemli.org>
patch2: https://lists.zx2c4.com/pipermail/cgit/2023-June/004843.html
Co-authored-by: Christian Hesse <mail@eworm.de>
Change-Id: I549a62e7c85c66d772edda997819a40f2d5835d7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8855
Autosubmit: sterni <sternenseemann@systemli.org>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Reviewed-by: grfn <grfn@gws.fyi>
This isn't necessary (it's all public!) and actually currently breaks
pulling from our public ECR repo
Change-Id: I32f0b92f5ca352f57a26d0e6c10ebf323aa006df
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8865
Reviewed-by: grfn <grfn@gws.fyi>
Autosubmit: grfn <grfn@gws.fyi>
Tested-by: BuildkiteCI
The issue this was working around has long since been fixed, and having
this disabled was preventing the service from working
Change-Id: I9b69f947b8952df786193f8784d0f5cc6b898440
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8862
Tested-by: BuildkiteCI
Autosubmit: grfn <grfn@gws.fyi>
Reviewed-by: grfn <grfn@gws.fyi>
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 runs a headscale server on sanduny which lets users join their
machines to the TVL tailscale network.
This would theoretically let people communicate with each other on the
internal network, but also more notably joined servers can advertise
exit node capability so that we can have our own "VPN network", for
starters with endpoints in Germany, UK and Russia (whitby, sanduny and
koptevo respectively).
This setup isn't fully stable yet, notably:
* The IP range used by tailscale is just the default one right now,
I'm not sure if that should be changed or what.
* The system is stateful (on sanduny), but the state is not (yet)
backed up anywhere. Use with caution.
* Machine joining is a manual process requiring SSH & root access to
sanduny.
The process is to log in to sanduny, then get a headscale shell with
`sudo -u headscale bash`, and to use the `headscale` CLI within
there to administrate access.
I've opted to create a user account `tvl` for TVL-owned machines,
and a personal account for myself and my machines.
Change-Id: I4f1be1fe8062a6c2e77203ff72fe8709f4e4dec8
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8837
Reviewed-by: sterni <sternenseemann@systemli.org>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
I've noticed this behavior when writing the admittedly cursed test case
included in this CL. Alternatively we could use some sort of machinery
using `builtins.trace`, but I don't think we capture stderr anywhere.
I've elected to put this into the eval cache itself while C++ Nix does
it in builtins.import already, namely via `realisePath`. We don't have
an equivalent for this yet, since we don't support any kind of IfD, but
we could revise that later. In any case, it seems good to encapsulate
`ImportCache` in this way, as it'll also allow using file hashes as
identifiers, for example.
C++ Nix also does our equivalent of canon_path in `builtins.import`
which we still don't, but I suspect it hardly makes a difference.
Change-Id: I05004737ca2458a4c67359d9e7d9a2f2154a0a0f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8839
Autosubmit: sterni <sternenseemann@systemli.org>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This machine is now being decomissioned.
Change-Id: Ib7f016c7de84dab2cdf3c071445cf830d2eccf5f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8838
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Also sets the `spdx` field which is specific to mkProvider and was
throwing eval warnings.
Change-Id: I37c04feb426b16f552fb0e0e2f188fd8d3bd0f03
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8836
Reviewed-by: sterni <sternenseemann@systemli.org>
Tested-by: BuildkiteCI
This will host a headscale server for TVL.
Change-Id: I8769852aaaf7a02a2d63f48ecf5adfd86747ff72
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8835
Reviewed-by: sterni <sternenseemann@systemli.org>
Tested-by: BuildkiteCI
This is a first implementation of a FUSE filesystem, mounting tvix-store
to a given location.
This is mostly meant as one additional lens into a store, and could be
used for builds. It's not meant to be used as a general-purpose thing.
It still has some rough edges:
- It doesn't implement open/close, so it doesn't use file handles.
Which means, we need to open blobs for partial reads over and over
again.
- It doesn't implement seek, as BlobReader doesn't implement seek yet.
- It doesn't track "lifetimes" of inodes by listening on forget,
meaning it might hold more data in memory than necessary.
- As we don't have store composition (and a caching layer) yet,
operations might be slow.
Change-Id: Ib1812ed761dfaf6aeb548443ae939c87530b7be8
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8667
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
When dealing with a formal argument in a function argument pattern that
has a default expression, there are two different things that can happen
at runtime: Either we select its value from the passed attribute
successfully or we need to use the default expression. Both of these may
be thunks and both of these may need finalisers. However, in the former
case this is taken care of elsewhere, the value will always be finalised
already if necessary. In the latter case we may need to finalise the
thunk resulting from the default expression. However, the thunk
corresponding to the expression may never end up in the local's stack
slot. Since finalisation goes by stack slot (and not constants), we need
to prevent a case where we don't fall back to the default expression,
but finalise anyways.
Previously, we worked around this by making `OpFinalise` ignore
non-thunks. Since finalisation of already evaluated thunks still
crashed, the faulty compilation of function pattern arguments could
still cause a crash.
As a new approach, we reinstate the old behavior of `OpFinalise` to
crash whenever encountering something that is either not a thunk or
doesn't need finalisation. This can also help catching (similar)
miscompilations in the future. To then prevent the crash, we need to
track whether we have fallen back or not at runtime. This is done using
an additional phantom on the stack that holds a new `FinaliseRequest`
value. When it comes to finalisation we check this value and
conditionally execute `OpFinalise` based on its value.
Resolves b/261 and b/265 (partially).
Change-Id: Ic04fb80ec671a2ba11fa645090769c335fb7f58b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8705
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Autosubmit: sterni <sternenseemann@systemli.org>
Pointed out by @Gaelan on GH in
https://github.com/tazjin/nix-1p/issues/12
As noted there, some Nix syntax highlighters will colour `import`
differently (e.g. the Emacs nix-mode), but it's not technically a
keyword in the language.
Change-Id: I1bbd36261cda31deb9ba47380a2810e64ba03ea1
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8832
Reviewed-by: sterni <sternenseemann@systemli.org>
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
The main instance is still running on polyanka, but things are moving
in this direction.
Change-Id: Idfa9e508023c05148003ac4621ae01dceb284c66
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8827
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
The GPT backend is cool, but it's also very slow, prone to request
errors and quite expensive.
This switches to Yandex Translate instead which for all posts that I
tested seems to be totally fine.
Change-Id: I5217113995b701508a83e7782eb1325957996719
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8826
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This adds a `from_str_with_config` function which takes a
user-supplied closure that sets additional settings on the
`tvix_eval::Evaluation`.
Note that users can not set `strict = false`, but other settings are
not restricted.
This solves b/262.
Change-Id: Ice184400b843cfbcaa5b6fe251ced12b6815e085
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8808
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
The entry list is now much more condensed. It's maybe a little *too*
condensed, but already closer to what I'm looking for.
Note: A new "note" post type has snuck in and can now be used for
random musings or comments on previous entries. Notes do not show up
in the Atom feed.
Change-Id: I920c0c7650937474b8a5f30cba78416554d523ce
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8806
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
