Commit graph

16333 commits

Author SHA1 Message Date
Robert Hensing
260c614762 Value: use std::span, change use of const
**`Value` and `const`**

These two deserve some explanation. We'll get to lists later.

Values can normally be thought of as immutable, except they are
are also the vehicle for call by need, which must be implemented
using mutation.

This circumstance makes a `const Value` a rather useless thing:

 - If it's a thunk, you can't evaluate it, except by copying, but
   that would not be call by need.

 - If it's not a thunk, you know the type, so the method that
   acquired it for you should have returned something more specific,
   such as a `const Bindings &` (which actually does make sense
   because that's an immutable span of pointers to mutable `Value`s.

 - If you don't care about the type yet, you might establish the
   convention that `const Value` means `deepSeq`-ed data, but
   this is hardly useful and not actually as safe as you would
   supposedly want to trust it to be - just convention.

**Lists**

`std::span` is a tuple of pointer and size - just what we need.

We don't return them as `const Value`, because considering the
first bullet point we discussed before, we'd have to force all
the list values, which isn't what we want.

So what we end up with is a nice representation of a list in
weak head normal form: the spine is immutable, but the
items may need some evaluation later.
2023-11-17 10:19:03 +01:00
Robert Hensing
7055c65285 Value: extract Value::Lambda 2023-11-17 10:19:03 +01:00
Robert Hensing
6af1d9f7b9 Value: extract Value::FunctionApplicationThunk 2023-11-17 10:19:03 +01:00
Robert Hensing
b55203e874 Value: extract Value::ClosureThunk 2023-11-17 10:19:03 +01:00
Robert Hensing
d8ff5cfe8e Value: extract Value::Path 2023-11-17 10:19:03 +01:00
Robert Hensing
2eb59c34b5 Value: extract Value::StringWithContext 2023-11-17 10:19:03 +01:00
Dominic Shelton
c819375769
doc: Add example of inherit in a let expression 2023-11-17 17:50:17 +11:00
Théophane Hufschmitt
7ba4e073e8
Merge pull request #6469 from gbpdt/fix/skipped_build_locking
Unlock output paths when a derivation is already built
2023-11-16 21:59:25 +01:00
Théophane Hufschmitt
fabae98ab4
Merge pull request #9360 from obsidiansystems/install-unit-tests
Allow installing unit tests
2023-11-16 21:25:21 +01:00
Théophane Hufschmitt
ac4431e9d0
Merge pull request #7348 from thufschmitt/dont-use-vlas
Remove the usage of VLAs in the code
2023-11-16 19:05:37 +01:00
Eelco Dolstra
57de482e00
Merge pull request #9364 from edolstra/fix-coverage
Fix a broken generated header file dependency
2023-11-16 17:37:05 +01:00
Eelco Dolstra
96d67620d5 Fix a broken generated header file dependency
https://hydra.nixos.org/build/240882042
2023-11-16 17:12:06 +01:00
Eelco Dolstra
31ebc6028b Fix symlink handling
This restores the symlink handling behaviour prior to
94812cca98.

Fixes #9298.
2023-11-16 16:45:14 +01:00
John Ericson
6c8f4ef350 Allow installing unit tests
Closes #9343

See that issue for motivation.

Installing these is disabled by default, but we enable it (and the
additional output we want isntall these too so as not to clutter the
existing ones) to use in cross builds and dev shells.
2023-11-16 09:55:42 -05:00
Robert Hensing
4e27f1947a libexpr: Reduce nonRecursiveStackReservation
128 is still beyond the point where the allocation overhead is
insignificant, but we don't anticipate to overflow for these
use cases, so it's fine.
2023-11-16 15:38:21 +01:00
Robert Hensing
a96be29db5 removeAttrs: increase stack reservation to 64 2023-11-16 15:38:21 +01:00
Robert Hensing
1b9813e4e6 primops: Name stack reservation limits 2023-11-16 15:38:21 +01:00
Robert Hensing
898c47384f primops: Err on the side of less stack usage
Try to stay away from stack overflows.

These small vectors use stack space. Most instances will not need
to allocate because in general most things are small, and large
things are worth heap allocating.

16 * 3 * word = 384 bytes is still quite a bit, but these functions
tend not to be part of deep recursions.
2023-11-16 15:38:21 +01:00
Robert Hensing
91114a6fa4 ExprCall::eval: Heap allocate at arity 5+ 2023-11-16 15:38:21 +01:00
Robert Hensing
206ece0f41 builtins.{any,all}: Use constant errorCtx
Clang warned that the expanded code used to have a buffer overflow.
Very strange, but also very avoidable.
2023-11-16 15:38:21 +01:00
Robert Hensing
9fa133dde5 readProcLink: Replace unnecessary value judgement by actual info 2023-11-16 15:38:21 +01:00
Robert Hensing
12c91a823e maxPrimOpArity: 64 -> 8
This makes stack usage significantly more compact, allowing larger
amounts of data to be processed on the same stack.

PrimOp functions with more than 8 positional (curried) arguments
should use an attrset instead.
2023-11-16 15:38:21 +01:00
Bob van der Linden
eaa3629532
fixup! nix: develop: add tests for interactive shell 2023-11-16 15:12:31 +01:00
Eelco Dolstra
2bd2702127
Merge pull request #9354 from vicky1999/fix/8875
Renamed nix nar dump-path to nix nar pack
2023-11-16 15:06:33 +01:00
Eelco Dolstra
16c052e4e7
Merge pull request #9307 from flox/tomberek.nix_key_newline
nix key: no need for progressBar
2023-11-16 15:05:34 +01:00
Eelco Dolstra
c47f769391
Merge pull request #9350 from NixOS/dependabot/github_actions/zeebe-io/backport-action-2.1.1
build(deps): bump zeebe-io/backport-action from 2.1.0 to 2.1.1
2023-11-16 15:01:49 +01:00
Robert Hensing
0daccb1121 libexpr: Check primop arity earlier 2023-11-16 12:28:32 +01:00
Théophane Hufschmitt
ba3cb4a049 Remove all the occurences of VLAs
There's generally no strict reason for using them, and they are somewhat
fishy, so let's avoid them.
2023-11-16 12:27:30 +01:00
Théophane Hufschmitt
5196613e82 Use boost small vectors instead of VLAs in the primops
VLAs are a dangerous feature, and their usage triggers an undefined
behavior since theire size can be zero in some cases.
So replace them with `boost::small_vector`s which fit the same goal but
are safer.

It's also incidentally consistently 1% faster on the benchmarks.
2023-11-16 12:27:30 +01:00
Bob van der Linden
ceab20d056
nix: develop: add tests for interactive shell 2023-11-16 09:37:49 +01:00
Bob van der Linden
8c54a01df5
nix: develop: always force SHELL to chosen shell
SHELL was inherited from the system environment. This resulted in a new
shell being started, but with SHELL still referring to the system shell
and not the one used by nix-develop.

Applications like make, use SHELL to run commands, which meant that
top-level commands are run inside the nix-develop-shell, but
sub-commands are ran inside the system shell.

This setenv forces SHELL to always be set to the shell used by
nix-develop.
2023-11-16 09:37:48 +01:00
John Ericson
e34c424279
Merge pull request #9357 from NixOS/nix-store-add
Add a new `nix store add` command
2023-11-15 13:49:44 -05:00
Théophane Hufschmitt
84128461b6 Add a new nix store add command
Deprecate `nix store add-file` and `nix store add-path`, and replace
them with a single `nix store add` command.
2023-11-15 19:21:17 +01:00
Eelco Dolstra
7ab91e7238 Implement shallow fetching 2023-11-15 15:00:17 +01:00
Eelco Dolstra
5dd4ae8687 Remove unused cacheType field 2023-11-15 14:08:34 +01:00
Eelco Dolstra
2890999911 Show Git fetch progress 2023-11-15 13:57:20 +01:00
Silvan Mosberger
70b396649c doc: logical implication is right-associative
nix-repl> bools = [ false true ]

    nix-repl> combinations = builtins.concatMap (a: builtins.concatMap (b: map (c: { inherit a b c; }) bools) bools) bools

    nix-repl> builtins.all ({ a, b, c }: (a -> b -> c) == (a -> (b -> c))) combinations
    true

    nix-repl> builtins.all ({ a, b, c }: (a -> b -> c) == ((a -> b) -> c)) combinations
    false
2023-11-15 07:04:09 +01:00
John Ericson
d23d0a074d
Merge pull request #9355 from obsidiansystems/fix-tiny-libutil-unit-test-make-bug
Fix makefile bug confusing `libnixutil-test` exe vs lib
2023-11-14 12:13:41 -05:00
John Ericson
9c7749e135 Fix makefile bug confusing libnixutil-test exe vs lib
The `-exe` variant is the program, the unsuffixed variant is the
library.

The corrected usage matches `libnixstore-test`.
2023-11-14 11:45:52 -05:00
Eelco Dolstra
2964a9f562 Fix relative submodule handling
Tested on

  nix flake prefetch 'git+https://github.com/blender/blender.git?rev=4ed8a360e956daf2591add4d3c9ec0719e2628fe&submodules=1'
2023-11-14 16:00:21 +01:00
Eelco Dolstra
6ec6b8aa36 Improve git submodule error reporting 2023-11-14 15:52:18 +01:00
vicky1999
e07e3c106a code cleanup 2023-11-14 20:02:33 +05:30
vicky1999
4944cdb94d nar dump-path command renamed to nar pack 2023-11-14 19:59:48 +05:30
Eelco Dolstra
c257c82447 Cleanup 2023-11-14 14:47:17 +01:00
Eelco Dolstra
7f576f5dfe Rename UnionInputAccessor to MountedInputAccessor 2023-11-14 14:02:57 +01:00
Eelco Dolstra
21140c987b Fix doxygen comments 2023-11-14 14:02:57 +01:00
Eelco Dolstra
4329bdf6a3 Move comment 2023-11-14 14:02:57 +01:00
Eelco Dolstra
25cf8f1071
src/libfetchers/union-input-accessor.cc: Apply suggestion
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
2023-11-14 13:57:24 +01:00
Eelco Dolstra
38b07d6347
src/libfetchers/git.cc: Apply suggestion
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
2023-11-14 13:38:03 +01:00
Eelco Dolstra
d74d2fdaa7 Move statusCallbackTrampoline 2023-11-14 13:35:26 +01:00