Directly fail if a flakeref points to something that isn't a directory
instead of falling back to the logic of trying to look up the hierarchy
to find a valid flake root.
Fix https://github.com/NixOS/nix/issues/9868
It is possible to exfiltrate a file descriptor out of the build sandbox
of FODs, and use it to modify the store path after it has been
registered.
To avoid that issue, don't register the output of the build, but a copy
of it (that will be free of any leaked file descriptor).
Instead, serialize as NAR and send that over, then rehash sever side.
This is alorithmically simpler, but comes at the cost of a newer
parameter to `Store::addToStoreFromDump`.
Co-authored-by: Eelco Dolstra <edolstra@gmail.com>
Part of RFC 133
Extracted from our old IPFS branches.
Co-Authored-By: Matthew Bauer <mjbauer95@gmail.com>
Co-Authored-By: Carlo Nucera <carlo.nucera@protonmail.com>
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
Co-authored-by: Florian Klink <flokli@flokli.de>
desugaring inherit-from to syntactic duplication of the source expr also
duplicates side effects of the source expr (such as trace calls) and
expensive computations (such as derivationStrict).
Flakes still reside in the Nix store (so there shouldn't be any change
in behaviour), but they are now accessed via the rootFS
accessor. Since rootFS implements access checks, we no longer have to
worry about flake.{nix,lock} or their parents being symlinks that
escape from the flake.
Extracted from the lazy-trees branch.
"hash type" -> "hash algorithm" in all comments, documentation, and
messages.
ht -> ha, [Hh]ashType -> [HhashAlgo] for all local variables and
function arguments. No API change is made.
Continuation of 5334c9c792 and 837b889c41.
The sandbox rule `(allow network* (local ip))` doesn't do what it
implies. Adding this rule permits all network traffic. We should be
matching on (remote ip "localhost:*")` instead.
This seems to have found one actual bug in fs-sink.cc: the symlink case
was falling into the regular file case, which can't possibly be
intentional, right?
When a file conflict arises during a package install a suggestion is
made to remove the old entry. This was previously done using the
installable URLs of the old entry. These URLs are quite verbose and
often do not equal the URL of the existing entry.
This change uses the recently introduced profile entry name for the
suggestion, resulting in a simpler output.
The improvement is easily seen in the change to the functional test.
- `nix store add` supports text hashing
With functional test ensuring it matches `builtins.toFile`.
- Factored-out flags for both commands
- Move all common reusable flags to `libcmd`
- They are not part of the *definition* of the CLI infra, just a usag
of it.
- The `libstore` flag couldn't go in `args.hh` in libutil anyways,
would be awkward for it to live alone
- Shuffle around `Cmd*` hierarchy so flags for deprecated commands don't
end up on the new ones
This PR reduces the creation of short-lived basic_json objects while
parsing flake.lock files. For large flake.lock files (~1.5MB) I was
observing ~60s being spent for trivial nix build operations while
after this change it is now taking ~1.6s.
It's better to just check whether the input has all the attributes
needed to consider itself locked (e.g. whether a Git input has an
'rev' attribute).
Also, the 'locked' field was actually incorrect for Git inputs: it
would be set to true even for dirty worktrees. As a result, we got
away with using fetchTree() internally even though fetchTree()
requires a locked input in pure mode. In particular, this allowed
'--override-input' to work by accident.
The fix is to pass a set of "overrides" to call-flake.nix for all the
unlocked inputs (i.e. the top-level flake and any --override-inputs).
This fixes warnings like
warning: Ignoring setting 'auto-allocate-uids' because experimental feature 'auto-allocate-uids' is not enabled
warning: Ignoring setting 'impure-env' because experimental feature 'configurable-impure-env' is not enabled
when using the daemon and the user didn't actually set those settings.
Note: this also hides those settings from `nix config show`, but that
seems a good thing.
`canonPath` and `absPath` work on native paths, and so should switch
between supporting Unix paths and Windows paths accordingly.
The templating is because `CanonPath`, which shares the implementation,
should always be Unix style. It is the pure "nix-native" path type for
virtual file operations --- it is part of Nix's "business logic", and
should not vary with the host OS accordingly.
The core `CanonPath` constructors were using `absPath`, but `absPath` in
some situations does IO which is not appropriate. It turns out that
these constructors avoided those situations, and thus were pure, but it
was far from obvious this was the case.
To remedy the situation, abstract the core algorithm from `canonPath` to
use separately in `CanonPath` without any IO. No we know by-construction
that those constructors are pure.
That leaves `CanonPath::fromCWD` as the only operation which uses IO /
is impure. Add docs on it, and `CanonPath` as a whole, explaining the
situation.
This is also necessary to support Windows paths on windows without
messing up `CanonPath`. But, I think it is good even without that.
Co-authored-by: Eelco Dolstra <edolstra@gmail.com>
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
There is no longer an `importTarball` method. Instead, there is a
`unpackTarfileToSink` function (back in libutil). The caller can use
thisw with the `getParseSink` method we added in the last commit easily
enough.
In addition, tarball cache functionality is separated from `git-utils`
and moved into `tarball-cache`. This ensures we are separating mechanism
and policy.
There is now a separation of:
1. A `FileSystemObjectSink` for writing to git repos
2. Adapting libarchive to use that parse sink.
The prepares a proper separation of concerns.
A command like
rm -rf ~/.cache/nix/tarball-cache/ ~/.cache/nix/fetcher-cache-v1.sqlite*; nix flake metadata 'git+file:///home/eelco/Dev/nixpkgs?rev=9463103069725474698139ab10f17a9d125da859'
was spending about 84% of its runtime in lookup(), specifically in
git_tree_entry_bypath(). (The reading of blobs is less than 3%.)
It appears libgit2 doesn't do a lot of caching of trees, so we now
make sure that when we look up a path, we add all its parents, and all
the immediate children of the parents (since we have them in memory
anyway), to our own cache.
This speed up the command above from 17.2s to 7.8s on my machine.
Fixes (or at least should improve a lot) #9684.
Commit 83c067c0fa changed `builtins.pathExists`
to resolve symlinks before checking for existence. Consequently, if the path
refers to a symlink itself, existence of the target of the symlink (instead of
the symlink itself) was checked. Restore the previous behavior by skipping
symlink resolution in the last component.
No outward facing behavior is changed.
Older methods with same names that operate on on method + algo pair (for
old-style `<method>:algo`) are renamed to `*WithAlgo`.)
The functions are unit-tested in the same way the names for the hash
algorithms are tested.
* reword description of the `cores` setting
- be precise about the `builder` executable
- clearly distinguish between `builder` and job parallelism
- clarify the role of `mkDerivation` in the example
- remove prose for the default, it's shown programmatically
- mention relation to `max-jobs`
- move all reference documentation to the `builders` configuration setting
- reword documentation on machine specification, add examples
- disable showing the default value, as it rendered as `@/dummy/machines`, which is wrong
- highlight the examples
- link to the configuration docs for distributed builds
- builder -> build machine
Co-authored-by: Janik H <janik@aq0.de>
for plain inherits this is really just a stylistic choice, but for
inherit-from it actually fixes an exponential size increase problem
during expr printing (as may happen during assertion failure reporting,
on during duplicate attr detection in the parser)
this also has the effect of sorting let bindings lexicographically
rather than by symbol creation order as was previously done, giving a
better canonicalization in the process.
When I started contributing to Nix, I found the mix of definitions and
names in `fmt.hh` to be rather confusing, especially the small
difference between `hintfmt` and `hintformat`. I've renamed many classes
and added documentation to most definitions.
- `formatHelper` is no longer exported.
- `fmt`'s documentation is now with `fmt` rather than (misleadingly)
above `formatHelper`.
- `yellowtxt` is renamed to `Magenta`.
`yellowtxt` wraps its value with `ANSI_WARNING`, but `ANSI_WARNING`
has been equal to `ANSI_MAGENTA` for a long time. Now the name is
updated.
- `normaltxt` is renamed to `Uncolored`.
- `hintfmt` has been merged into `hintformat` as extra constructor
functions.
- `hintformat` has been renamed to `hintfmt`.
- The single-argument `hintformat(std::string)` constructor has been
renamed to a static member `hintformat::interpolate` to avoid pitfalls
with using user-generated strings as format strings.
As discussed in the last Nix team meeting (2024-02-95), this method
doesn't belong because `CanonPath` is a virtual/ideal absolute path
format, not used in file systems beyond the native OS format for which a
"current working directory" is defined.
Progress towards #9205
Pretty-print values in the REPL by printing each item in a list or
attrset on a separate line. When possible, single-item lists and
attrsets are printed on one line, as long as they don't contain a nested
list, attrset, or thunk.
Before:
```
{ attrs = { a = { b = { c = { }; }; }; }; list = [ 1 ]; list' = [ 1 2 3 ]; }
```
After:
```
{
attrs = {
a = {
b = {
c = { };
};
};
};
list = [ 1 ];
list' = [
1
2
3
];
}
```
Some tools which consume the "nix print-dev-env" rc script (such as
"nix-direnv") are sensitive to the use of unbound variables. They use
"set -u".
The "nix print-dev-env" rc script initially unsets "shellHook", then
loads variables from the derivation, and then evaluates "shellHook".
However, most derivations don't have a "shellHook" attribute.
So users get the error "shellHook: unbound variable". This can be
demonstrated with the command:
nix print-dev-env nixpkgs#hello | bash -u
This commit changes the rc script to provide an empty fallback value
for the "shellHook" variable.
Closes: #7951#8253
It is entirely possible for the path to be an empty string and many
unit tests actually pass it as an empty string (e.g. both_roundrip or
turnsEmptyPathIntoCWD). In this case, without this patch, absPath will
perform a one-byte out-of-bounds access.
This was discovered while enabling the nix test suite on Alpine where
we compile all software with `-D_GLIBCXX_ASSERTIONS=1`, thus resulting
in a test failure on Alpine.
While preparing PRs like #9753, I've had to change error messages in
dozens of code paths. It would be nice if instead of
EvalError("expected 'boolean' but found '%1%'", showType(v))
we could write
TypeError(v, "boolean")
or similar. Then, changing the error message could be a mechanical
refactor with the compiler pointing out places the constructor needs to
be changed, rather than the error-prone process of grepping through the
codebase. Structured errors would also help prevent the "same" error
from having multiple slightly different messages, and could be a first
step towards error codes / an error index.
This PR reworks the exception infrastructure in `libexpr` to
support exception types with different constructor signatures than
`BaseError`. Actually refactoring the exceptions to use structured data
will come in a future PR (this one is big enough already, as it has to
touch every exception in `libexpr`).
The core design is in `eval-error.hh`. Generally, errors like this:
state.error("'%s' is not a string", getAttrPathStr())
.debugThrow<TypeError>()
are transformed like this:
state.error<TypeError>("'%s' is not a string", getAttrPathStr())
.debugThrow()
The type annotation has moved from `ErrorBuilder::debugThrow` to
`EvalState::error`.
As discussed in the maintainer meeting on 2024-01-29.
Mainly this is to avoid a situation where the name is parsed and
treated as a file name, mostly to protect users.
.-* and ..-* are also considered invalid because they might strip
on that separator to remove versions. Doesn't really work, but that's
what we decided, and I won't argue with it, because .-* probably
doesn't seem to have a real world application anyway.
We do still permit a 1-character name that's just "-", which still
poses a similar risk in such a situation. We can't start disallowing
trailing -, because a non-zero number of users will need it and we've
seen how annoying and painful such a change is.
What matters most is preventing a situation where . or .. can be
injected, and to just get this done.
To quote the method doc:
Non-impure derivations can still behave impurely, to the degree permitted
by the sandbox. Hence why this method isn't `isPure`: impure derivations
are not the negation of pure derivations. Purity can not be ascertained
except by rather heavy tools.
The code works fine on macOS, but the default stack size we attempt to
set is larger than what my system will allow (Nix attempts to set the
stack size to 67108864, but the maximum allowed is 67092480), so I've
instead used the requested stack size or the maximum allowed, whichever
is smaller.
I've also added an error message if setting the stack size fails. It
looks like this:
> Failed to increase stack size from 8372224 to 67108864 (maximum
> allowed stack size: 67092480): Invalid argument
This extends the `error: cannot coerce a TYPE to a string` message
to print the value that could not be coerced. This helps with debugging
by making it easier to track down where the value is being produced
from, especially in errors with deep or unhelpful stack traces.
This is more conceptually correct (the order does not matter), and also
matches what Hydra already does.
(Nix and Hydra matching is needed for dedup
https://github.com/NixOS/hydra/issues/1164)
More invariants are enforced in the type, and less state needs to be
stored in the main sink itself. The method here is roughly that known as
"session types".
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
This avoids split-on-whitespace errors:
- No more `bash -c` needed
- No more `shellEscape` needed
- `remote-program` ssh store setting also cleanly supports args (e.g.
`nix daemon`)
- `ssh` uses `--` to separate args for SSH from args for the command to
run.
and will help with Hydra dedup.
Some code taken from #6628.
Co-Authored-By: Alexander Bantyev <balsoft@balsoft.ru>
Low-hanging fruit in the spirit of #9753 and #9754 (means 9999years did
all the hard work already).
This basically prints out what was attempted to be called as function,
i.e.
map (import <nixpkgs> {}) [ 1 2 3 ]
now gives the following error message:
error:
… while calling the 'map' builtin
at «string»:1:1:
1| map (import <nixpkgs> {}) [ 1 2 3 ]
| ^
… while evaluating the first argument passed to builtins.map
error: expected a function but found a set: { _type = "pkgs"; AAAAAASomeThingsFailToEvaluate = «thunk»; AMB-plugins = «thunk»; ArchiSteamFarm = «thunk»; BeatSaberModManager = «thunk»; CHOWTapeModel = «thunk»; ChowCentaur = «thunk»; ChowKick = «thunk»; ChowPhaser = «thunk»; CoinMP = «thunk»; «18783 attributes elided»}
Factor out `ServeProto::BasicClientConnection` for Hydra to share
- `queryValidPaths`: Hydra uses the lock argument differently than Nix,
so we un-hard-code it.
- `buildDerivationRequest`: Just the request half, as Hydra does some
things between requesting and responding.
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
Do this if we want to do `--hash-algo` everywhere, and not `--algo` for
hash commands.
The new `nix hash convert` is updated. Deprecated new CLI commands are
left as-is (`nix hash path` needs to be redone and is also left as-is).
Good to document these formats separately from commands that happen to
use them.
Eventually I would like this and `builtins.derivation` to refer to a
store section on derivations that is authoritative, but that doesn't yet
exist, and will take some time to make. So I think we're just best off
merging this now as is.
Co-authored-by: Valentin Gagarin <valentin.gagarin@tweag.io>
Return a value instead of throwing.
Rather than the more trivial refactor of wrapping the return value in
another std::optional, we retain the meaning of the outer optional:
"we know at least something."
So we have changed:
return nullopt -> return nullopt
throw InvalidPath -> return make_optional(nullptr)
return vpi -> return make_optional(vpi)
Add several tests for git fetching:
- shallow-cache-separation: can fetch the same repo shallowly and non-shallowly
- shallow-ignore-ref: ensure that ref gets ignored when shallow=true is set
- ssh-shallow: can fetch a git repo via ssh using shallow=1
libgit2 is not capable of using git-credentials helpers yet.
This prevents private repositories from being used.
Based on code that was replaced in https://github.com/NixOS/nix/pull/9240
(Introduce libgit2); hence:
Co-authored-by: Eelco Dolstra <edolstra@gmail.com>
In rare cases (e.g. when using allowSubstitutes = false), it's
possible that we simultaneously have a DerivationGoal *and* a
SubstitutionGoal building the same path. So if a DerivationGoal
already built the path while the SubstitutionGoal was waiting for a
download slot, it saves us a superfluous download to exit early.
In the "discard" case (i.e. when the store path already exists
locally), when we call parseDump() from a Finally and it throws an
exception (e.g. if the download of the NAR fails), Nix crashes:
terminate called after throwing an instance of 'nix::SubstituteGone'
what(): error: file 'nar/06br3254rx4gz4cvjzxlv028jrx80zg5i4jr62vjmn416dqihgr7.nar.xz' does not exist in binary cache 'http://localhost'
Aborted (core dumped)
Instead of having it be the default method in `Store` itself, have it be
the implementation in `DummyStore` and `LegacySSHStore`. Then just the
implementations which fail to provide the method pay the "penalty" of
dealing with the icky `unimplemented` function for non-compliance.
Picks up where #8217. Getting close to no `unsupported` in the `Store`
interface itself!
More progress on issue #5729.
It is good to propagate the underlying error so whether or not we use a
process to deal with path length issues is not observable.
Also, as these wrapper functions got more and more complex, the code
duplication got worse and worse. The new `bindConnectProcHelper`
function deduplicates them.
This is useful for determining quickly which substituters to query.
An alternative would be for users to invoke the narinfo cache db directly,
so why do we need this change?
- It is easier to use. I believe Nix itself should also use it.
- This way, the narinfo cache db remains an implementation detail.
- Callers get to use the in-memory cache as well.
This does not yet resolve the coupling between packages and
derivations, but it makes the code more consistent with the
terminology, and it accentuates places where the coupling is
obvious, such as
auto drvPath = packageInfo.queryDrvPath();
if (!drvPath)
throw Error("'%s' is not a derivation", what());
... which isn't wrong, and in my opinion, doesn't even look
wrong, because it just reflects the current logic.
However, I do like that we can now start to see in the code that
this coupling is perhaps a bit arbitrary.
After this rename, we can bring the DerivingPath concept into type
and start to lift this limitation.
these symbols are used a *lot*, so it makes sense to cache them. this
mostly increases clarity of the code (however clear one may wish to call
the parser desugaring here), but it also provides a small performance
benefit.
there's no reason the parser itself should be doing semantic analysis
like bindVars. split this bit apart (retaining the previous name in
EvalState) and have the parser really do *only* parsing, decoupled from
EvalState.
most EvalState and Expr members defined here could be elsewhere, where
they'd be easier to maintain (not being embedded in a file with arcane
syntax) and *somewhat* more faithfully placed according to the path of
the file they're defined in.
most instances of this being used do not refer to the "current"
position, sometimes not even to one reasonably close by. it could also
be called `makePos` instead, but `at` seems clear in context.
ParserState better describes what this struct really is. the parser
really does modify its state (most notably position and symbol tables),
so calling it that rather than obliquely "data" (which implies being
input only) makes sense.
since nix doesn't use the bison `error` terminal anywhere any invocation
of yyerror will immediately cause a failure. since we're *already*
leaking tons of memory whatever little bit bison allocates internally
doesn't much matter any more, and we'll be replacing the parser soon anyway.
coincidentally this now also matches the error behavior of URIs when
they are disabled or ~/ paths in pure eval mode, duplicate attr
detection etc.
The data was (accidentally?) copied into a std::string,
even though the string is immediately converted into a std::string_view.
The code has been changed to construct a std::string_view directly,
such that one copy less happens.
`FLOAT`, `INT`, and `IN` are identifers taken by macros.
The name `IN_KW` is chosen to match `OR_KW`, which is presumably named
that way for the same reason of dodging macros.
Most of this is a `catch SysError` -> `catch SystemError` sed. This
is a rather pure-churn change I would like to get out of the way. **The
intersting part is `src/libutil/error.hh`.**
On Unix, we will only throw the `SysError` concrete class, which has
the same constructors that `SystemError` used to have.
On Windows, we will throw `WinError` *and* `SysError`. `WinError`
(which will be created in a later PR), will use a `DWORD` instead of
`int` error value, and `GetLastError()`, which is the Windows equivalent
of the `errno` machinery. Windows will *also* use `SysError` because
Window's "libc" (MSVCRT) implements the POSIX interface, and we use it
too.
As the docs describe, while we *throw* one of the 3 choices above (2
concrete classes or the alias), we should always *catch* `SystemError`.
This ensures no matter how the implementation changes for Windows (e.g.
between `SysError` and `WinError`) the catching logic stays the same
and stays correct.
Co-Authored-By volth <volth@volth.com>
Co-Authored-By Eugene Butler <eugene@eugene4.com>
When returning a 0-length substring, avoid calling coerceToString,
since it returns a string_view with the string's length, which is
expensive to compute for large strings.
Also fingerprint and some preparatory improvements.
Testing is still not up to scratch because lots of logic is duplicated
between the workdir and commit cases.
Enabled for fetchGit, which historically had this behavior,
among other behaviors we do not want in fetchGit.
fetchTree disables this parameter by default. It can choose the
simpler behavior, as it is still experimental.
I am not confident that the filtering implementation is future
proof. It should reuse a source filtering wrapper, which I believe
Eelco has already written, but not merged yet.
The Nix team has requested that this output format remain unchanged.
I've added a warning to the man page explaining that `nix-instantiate
--eval` output will not parse correctly in many situations.
Previously, there were two mostly-identical value printers -- one in
`libexpr/eval.cc` (which didn't force values) and one in
`libcmd/repl.cc` (which did force values and also printed ANSI color
codes).
This PR unifies both of these printers into `print.cc` and provides a
`PrintOptions` struct for controlling the output, which allows for
toggling whether values are forced, whether repeated values are tracked,
and whether ANSI color codes are displayed.
Additionally, `PrintOptions` allows tuning the maximum number of
attributes, list items, and bytes in a string that will be displayed;
this makes it ideal for contexts where printing too much output (e.g.
all of Nixpkgs) is distracting. (As requested by @roberth in
https://github.com/NixOS/nix/pull/9554#issuecomment-1845095735)
Please read the tests for example output.
Future work:
- It would be nice to provide this function as a builtin, perhaps
`builtins.toStringDebug` -- a printing function that never fails would
be useful when debugging Nix code.
- It would be nice to support customizing `PrintOptions` members on the
command line, e.g. `--option to-string-max-attrs 1000`.
Changes:
- CPP variable is now `USE_READLINE` not `READLINE`
- `configure.ac` supports with new CLI flag
- `package.nix` supports with new configuration option
- `flake.nix` CIs this (along with no markdown)
Remove old Ubuntu 16.04 stop-gap too, as that is now quite old.
Motivation:
- editline does not build for Windows, but readline *should*. (I am
still working on this in Nixpkgs at this time, however. So there will
be a follow-up Nix PR removing the windows-only skipping of the
readline library once I am done.)
- Per
https://salsa.debian.org/debian/nix/-/blob/master/debian/rules?ref_type=heads#L27
and #2551, Debian builds Nix with readline. Now we better support and
CI that build configuration.
This is picking up where #2551 left off, ensuring we test a few more
things not merely have CPP for them.
Co-authored-by: Weijia Wang <9713184+wegank@users.noreply.github.com>
Also move `SourcePath` into `libutil`.
These changes allow `error.hh` and `error.cc` to access source path and
position information, which we can use to produce better error messages
(for example, we could consider omitting filenames when two or more
consecutive stack frames originate from the same file).
This sets up infrastructure in libutil to allow for signing other than
by a secret key in memory. #9076 uses this to implement remote signing.
(Split from that PR to allow reviewing in smaller chunks.)
Co-Authored-By: Raito Bezarius <masterancpp@gmail.com>
This avoids a Value allocation for empty list constants. During a `nix
search nixpkgs`, about 82% of all thunked lists are empty, so this
removes about 3 million Value allocations.
Performance comparison on `nix search github:NixOS/nixpkgs/e1fa12d4f6c6fe19ccb59cac54b5b3f25e160870 --no-eval-cache`:
maximum RSS: median = 3845432.0000 mean = 3845432.0000 stddev = 0.0000 min = 3845432.0000 max = 3845432.0000 [rejected?, p=0.00000, Δ=-70084.00000±0.00000]
soft page faults: median = 965395.0000 mean = 965394.6667 stddev = 1.1181 min = 965392.0000 max = 965396.0000 [rejected?, p=0.00000, Δ=-17929.77778±38.59610]
system CPU time: median = 1.8029 mean = 1.7702 stddev = 0.0621 min = 1.6749 max = 1.8417 [rejected, p=0.00064, Δ=-0.12873±0.09905]
user CPU time: median = 14.1022 mean = 14.0633 stddev = 0.1869 min = 13.8118 max = 14.3190 [not rejected, p=0.03006, Δ=-0.18248±0.24928]
elapsed time: median = 15.8205 mean = 15.8618 stddev = 0.2312 min = 15.5033 max = 16.1670 [not rejected, p=0.00558, Δ=-0.28963±0.29434]
since `up` and `values` are both pointer-aligned the type field will
also be pointer-aligned, wasting 48 bits of space on most machines. we
can get away with removing the type field altogether by encoding some
information into the `with` expr that created the env to begin with,
reducing the GC load for the absolutely massive amount of single-entry
envs we create for lambdas. this reduces memory usage of system eval by
quite a bit (reducing heap size of our system eval from 8.4GB to 8.23GB)
and gives similar savings in eval time.
running `nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'`
before:
Time (mean ± σ): 5.576 s ± 0.003 s [User: 5.197 s, System: 0.378 s]
Range (min … max): 5.572 s … 5.581 s 10 runs
after:
Time (mean ± σ): 5.408 s ± 0.002 s [User: 5.019 s, System: 0.388 s]
Range (min … max): 5.405 s … 5.411 s 10 runs
many paths need not be heap-allocated, and derivation env name/valye
pairs can be moved into the map.
before:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.883 s ± 0.016 s [User: 5.250 s, System: 1.424 s]
Range (min … max): 6.860 s … 6.905 s 10 runs
after:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.868 s ± 0.027 s [User: 5.194 s, System: 1.466 s]
Range (min … max): 6.828 s … 6.913 s 10 runs
the table is very small compared to cache sizes and a single indexed
load is faster than three comparisons.
before:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.907 s ± 0.012 s [User: 5.272 s, System: 1.429 s]
Range (min … max): 6.893 s … 6.926 s 10 runs
after:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.883 s ± 0.016 s [User: 5.250 s, System: 1.424 s]
Range (min … max): 6.860 s … 6.905 s 10 runs
a bunch of derivation strings contain no escape sequences. we can
optimize for this fact by first scanning for the end of a derivation
string and simply returning the contents unmodified if no escape
sequences were found. to make this even more efficient we can also use
BackedStringViews to avoid copies, avoiding heap allocations for
transient data.
before:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.952 s ± 0.015 s [User: 5.294 s, System: 1.452 s]
Range (min … max): 6.926 s … 6.974 s 10 runs
after:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.907 s ± 0.012 s [User: 5.272 s, System: 1.429 s]
Range (min … max): 6.893 s … 6.926 s 10 runs
This fixes a segfault on infinite function call recursion (rather than
infinite thunk recursion) by tracking the function call depth in
`EvalState`.
Additionally, to avoid printing extremely long stack traces, stack
frames are now deduplicated, with a `(19997 duplicate traces omitted)`
message. This should only really be triggered in infinite recursion
scenarios.
Before:
$ nix-instantiate --eval --expr '(x: x x) (x: x x)'
Segmentation fault: 11
After:
$ nix-instantiate --eval --expr '(x: x x) (x: x x)'
error: stack overflow
at «string»:1:14:
1| (x: x x) (x: x x)
| ^
$ nix-instantiate --eval --expr '(x: x x) (x: x x)' --show-trace
error:
… from call site
at «string»:1:1:
1| (x: x x) (x: x x)
| ^
… while calling anonymous lambda
at «string»:1:2:
1| (x: x x) (x: x x)
| ^
… from call site
at «string»:1:5:
1| (x: x x) (x: x x)
| ^
… while calling anonymous lambda
at «string»:1:11:
1| (x: x x) (x: x x)
| ^
… from call site
at «string»:1:14:
1| (x: x x) (x: x x)
| ^
(19997 duplicate traces omitted)
error: stack overflow
at «string»:1:14:
1| (x: x x) (x: x x)
| ^
more buffers that can be uninitialized and on the stack. small
difference, but still worth doing.
before:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.963 s ± 0.011 s [User: 5.330 s, System: 1.421 s]
Range (min … max): 6.943 s … 6.974 s 10 runs
after:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.952 s ± 0.015 s [User: 5.294 s, System: 1.452 s]
Range (min … max): 6.926 s … 6.974 s 10 runs
istream sentry objects are very expensive for single-character
operations, and since we don't configure exception masks for the
istreams used here they don't even do anything. all we need is
end-of-string checks and an advancing position in an immutable memory
buffer, both of which can be had for much cheaper than istreams allow.
the effect of this change is most apparent on empty stores.
before:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 7.167 s ± 0.013 s [User: 5.528 s, System: 1.431 s]
Range (min … max): 7.147 s … 7.182 s 10 runs
after:
Benchmark 1: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 6.963 s ± 0.011 s [User: 5.330 s, System: 1.421 s]
Range (min … max): 6.943 s … 6.974 s 10 runs
Previously, IFDs would be built within the eval store, even though one
is typically using `--eval-store` precisely to *avoid* local builds.
Because the resulting Nix expression must be copied back to the eval
store in order to be imported, this requires the eval store to trust
the build store's signatures.
The profile manifest is now an object keyed on the name returned by
getNameFromURL() at installation time, instead of an array. This
ensures that the names of profile elements don't change when other
elements are added/removed.
Prior to this change, Nix would prepend every installable to the PATH
list in order to ensure that installables appeared before the current
PATH from the ambient environment.
With this change, all the installables are still prepended to the PATH,
but in the same order as they appear on the command line. This means
that the first of two packages that expose an executable `hello` would
appear in the PATH first, and thus be executed first.
See the test in the prior commit for a more concrete example.
There's no good reason to deprecate it:
- For consistency reasons it should continue to exist, such that all
primitive types have a corresponding `builtins.is*` primop.
- There's no implementation cost to continuing to have this function
- It costs users time to try to migrate away from it, e.g.
https://github.com/NixOS/nixpkgs/pull/219747 and https://github.com/NixOS/nixpkgs/pull/275548
- Using it can give easier-to-read code like `all isNull list`
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
this also reduces forceValue code size and removes the need for
hideInDiagnostics. coopting thunk forcing like this has the additional
benefit of clarifying how these errors can happen in the first place.
forceValue is extremely hot. interestingly adding likeliness annotations
to the branches does not seem to make a difference.
before:
Time (mean ± σ): 4.224 s ± 0.005 s [User: 3.711 s, System: 0.512 s]
Range (min … max): 4.218 s … 4.234 s 10 runs
after:
Time (mean ± σ): 4.140 s ± 0.009 s [User: 3.647 s, System: 0.492 s]
Range (min … max): 4.130 s … 4.152 s 10 runs
almost all uses of this are interactive, except for deepSeq. deepSeq is
going to be expensive and rare enough to not care much about, and
Value::determinePos should usually be cheap enough to not be too much of
a burden in any case.
~1% parser speedup from not using TLS indirections, less on system eval.
this could have also gone in flex yyextra data, but that's significantly
slower for some reason (albeit still faster than thread locals).
before:
Time (mean ± σ): 4.231 s ± 0.004 s [User: 3.725 s, System: 0.504 s]
Range (min … max): 4.226 s … 4.240 s 10 runs
after:
Time (mean ± σ): 4.224 s ± 0.005 s [User: 3.711 s, System: 0.512 s]
Range (min … max): 4.218 s … 4.234 s 10 runs
~2% speedup on parsing without eval, less (but still significant) on
system eval. having flex generate faster parsers leads to very strange
misparses. maybe re2c is worth investigating.
before:
Time (mean ± σ): 4.260 s ± 0.003 s [User: 3.754 s, System: 0.505 s]
Range (min … max): 4.257 s … 4.266 s 10 runs
after:
Time (mean ± σ): 4.231 s ± 0.004 s [User: 3.725 s, System: 0.504 s]
Range (min … max): 4.226 s … 4.240 s 10 runs
as written the comparisons generate copies, even though it looks as
though they shouldn't.
before:
Time (mean ± σ): 4.396 s ± 0.002 s [User: 3.894 s, System: 0.501 s]
Range (min … max): 4.393 s … 4.399 s 10 runs
after:
Time (mean ± σ): 4.260 s ± 0.003 s [User: 3.754 s, System: 0.505 s]
Range (min … max): 4.257 s … 4.266 s 10 runs
