When used with `readFile`, we have a pretty good heuristic of the file
size, so `reserve` this in the `string`. This will save some allocation
/ copy when the string is growing.
The commit 3cc1125595 adds a `grantpt`
call on the builder pseudo terminal fd. This call is actually only
required for MacOS, but it however requires a RW access to /dev/pts
which is only RO bindmounted in the Bazel Linux sandbox. So, Nix can
not be actually run in the Bazel Linux sandbox for unneeded reasons.
This closes#3026 by allowing `builtins.readFile` to read a file with a
wrongly reported file size, for example, files in `/proc` may report a
file size of 0. Reading file in `/proc` is not a good enough motivation,
however I do think it just makes nix more robust by allowing more file
to be read. Especially, I do considerer the previous behavior to be
dangerous because nix was previously reading truncated files. Examples
of file system which incorrectly report file size may be network file
system or dynamic file system (for performance reason, a dynamic file
system such as FUSE may generate the content of the file on demand).
```
nix-repl> builtins.readFile "/proc/version"
""
```
With this commit:
```
nix-repl> builtins.readFile "/proc/version"
"Linux version 5.6.7 (nixbld@localhost) (gcc version 9.3.0 (GCC)) #1-NixOS SMP Thu Apr 23 08:38:27 UTC 2020\n"
```
Here is a summary of the behavior changes:
- If the reported size is smaller, previous implementation
was silently returning a truncated file content. The new implementation
is returning the correct file content.
- If a file had a bigger reported file size, previous implementation was
failing with an exception, but the new implementation is returning the
correct file content. This change of behavior is coherent with this pull
request.
Open questions
- The behavior is unchanged for correctly reported file size, however
performances may vary because it uses the more complex sink interface.
Considering that sink is used a lot, I don't think this impacts the
performance a lot.
- `builtins.readFile` on an infinite file, such as `/dev/random` may
fill the memory.
- it does not support adding file to store, such as `${/proc/version}`.
Suppose I have a path /nix/store/[hash]-[name]/a/a/a/a/a/[...]/a,
long enough that everything after "/nix/store/" is longer than 4096
(MAX_PATH) bytes.
Nix will happily allow such a path to be inserted into the store,
because it doesn't look at all the nested structure. It just cares
about the /nix/store/[hash]-[name] part. But, when the path is deleted,
we encounter a problem. Nix will move the path to /nix/store/trash, but
then when it's trying to recursively delete the trash directory, it will
at some point try to unlink
/nix/store/trash/[hash]-[name]/a/a/a/a/a/[...]/a. This will fail,
because the path is too long. After this has failed, any store deletion
operation will never work again, because Nix needs to delete the trash
directory before recreating it to move new things to it. (I assume this
is because otherwise a path being deleted could already exist in the
trash, and then moving it would fail.)
This means that if I can trick somebody into just fetching a tarball
containing a path of the right length, they won't be able to delete
store paths or garbage collect ever again, until the offending path is
manually removed from /nix/store/trash. (And even fixing this manually
is quite difficult if you don't understand the issue, because the
absolute path that Nix says it failed to remove is also too long for
rm(1).)
This patch fixes the issue by making Nix's recursive delete operation
use unlinkat(2). This function takes a relative path and a directory
file descriptor. We ensure that the relative path is always just the
name of the directory entry, and therefore its length will never exceed
255 bytes. This means that it will never even come close to AX_PATH,
and Nix will therefore be able to handle removing arbitrarily deep
directory hierachies.
Since the directory file descriptor is used for recursion after being
used in readDirectory, I made a variant of readDirectory that takes an
already open directory stream, to avoid the directory being opened
multiple times. As we have seen from this issue, the less we have to
interact with paths, the better, and so it's good to reuse file
descriptors where possible.
I left _deletePath as succeeding even if the parent directory doesn't
exist, even though that feels wrong to me, because without that early
return, the linux-sandbox test failed.
Reported-by: Alyssa Ross <hi@alyssa.is>
Thanks-to: Puck Meerburg <puck@puckipedia.com>
Tested-by: Puck Meerburg <puck@puckipedia.com>
Reviewed-by: Puck Meerburg <puck@puckipedia.com>
Reduces the number of store queries it performs. Also prints a warning
if any of the selectors did not match any installed derivations.
UX Caveats:
- Will print a warning that nothing matched if a previous selector
already removed the path
- Will not do anything if no selectors were provided (no change from
before).
Fixes#3531
Usually this just writes to stdout, but for ProgressBar, we need to
clear the current line, write the line to stdout, and then redraw the
progress bar.
(cherry picked from commit 696c026006)
Motivation: maintain project-level configuration files.
Document the whole situation a bit better so that it corresponds to the
implementation, and add NIX_USER_CONF_FILES that allows overriding
which user files Nix will load during startup.
Previously the memory would occasionally be collected during eval since
the GC doesn't consider the member variable as alive / doesn't scan the
region of memory where the pointer lives.
By using the traceable_allocator<T> allocator provided by Boehm GC we
can ensure the memory isn't collected. It should be properly freed when
SourceExprCommand goes out of scope.
This doesn't just cause problems for nix-store --serve but also results
in certain build failures. Builds that use unix domain sockets in their
tests often fail because the /var/folders prefix already consumes more
than half of the maximum length of socket paths.
struct sockaddr_un {
sa_family_t sun_family; /* AF_UNIX */
char sun_path[108]; /* Pathname */
};
Temporarily add user-write permission to build directory so that it
can be moved out of the sandbox to the store with a .check suffix.
This is necessary because the build directory has already had its
permissions set read-only, but write permission is required
to update the directory's parent link to move it out of the sandbox.
Updated the related --check "derivation may not be deterministic"
messages to consistently use the real store paths.
Added test for non-root sandbox nix-build --check -K to demonstrate
issue and help prevent regressions.
With --check and the --keep-failed (-K) flag, the temporary directory
was being retained regardless of whether the build was successful and
reproducible. This removes the temporary directory, as expected, on
a reproducible check build.
Added tests to verify that temporary build directories are not
retained unnecessarily, particularly when using --check with
--keep-failed.
This fetchers copies a plain directory (i.e. not a Git/Mercurial
repository) to the store (or does nothing if the path is already a
store path).
One use case is to pin the 'nixpkgs' flake used to build the current
NixOS system, and prevent it from being garbage-collected, via a
system registry entry like this:
{
"from": {
"id": "nixpkgs",
"type": "indirect"
},
"to": {
"type": "path",
"path": "/nix/store/rralhl3wj4rdwzjn16g7d93mibvlr521-source",
"lastModified": 1585388205,
"rev": "b0c285807d6a9f1b7562ec417c24fa1a30ecc31a"
},
"exact": true
}
Note the fake "lastModified" and "rev" attributes that ensure that the
flake gives the same evaluation results as the corresponding
Git/GitHub inputs.
(cherry picked from commit 12f9379123)
This provides a pluggable mechanism for defining new fetchers. It adds
a builtin function 'fetchTree' that generalizes existing fetchers like
'fetchGit', 'fetchMercurial' and 'fetchTarball'. 'fetchTree' takes a
set of attributes, e.g.
fetchTree {
type = "git";
url = "https://example.org/repo.git";
ref = "some-branch";
rev = "abcdef...";
}
The existing fetchers are just wrappers around this. Note that the
input attributes to fetchTree are the same as flake input
specifications and flake lock file entries.
All fetchers share a common cache stored in
~/.cache/nix/fetcher-cache-v1.sqlite. This replaces the ad hoc caching
mechanisms in fetchGit and download.cc (e.g. ~/.cache/nix/{tarballs,git-revs*}).
This also adds support for Git worktrees (c169ea5904).
When encountering an unsupported protocol, there's no need to retry.
Chances are, it won't suddenly be supported between retry attempts;
error instead. Otherwise, you see something like the following:
$ nix-env -i -f git://git@github.com/foo/bar
warning: unable to download 'git://git@github.com/foo/bar': Unsupported protocol (1); retrying in 335 ms
warning: unable to download 'git://git@github.com/foo/bar': Unsupported protocol (1); retrying in 604 ms
warning: unable to download 'git://git@github.com/foo/bar': Unsupported protocol (1); retrying in 1340 ms
warning: unable to download 'git://git@github.com/foo/bar': Unsupported protocol (1); retrying in 2685 ms
With this change, you now see:
$ nix-env -i -f git://git@github.com/foo/bar
error: unable to download 'git://git@github.com/foo/bar': Unsupported protocol (1)
This allows querying the location of function arguments. E.g.
builtins.unsafeGetAttrPos "x" (builtins.functionArgs ({ x }: null))
=> { column = 57; file = "/home/infinisil/src/nix/inst/test.nix"; line = 1; }
Using std::filesystem means also having to link with -lstdc++fs on
some platforms and it's hard to discover for what platforms this is
needed. As all the functionality is already implemented as utilities,
use those instead.
Due to fetchGit not checking if rev is an ancestor of ref (there is even
a FIXME comment about it in the code), the cache repo might not have the
ref even though it has the rev. This doesn't matter when submodule =
false, but the submodule = true code blows up because it tries to fetch
the (missing) ref from the cache repo.
Fix this in the simplest way possible: fetch all refs from the local
cache repo when submodules = true.
TODO: Add tests.
The .link file is used as a lock, so I think we should put the
"submodule" attribute in there since turning on submodules creates a new
.link file path.
Sadly 10.15 changed /bin/sh to a shim which executes bash, this means it
can't be used anymore without also opening up the sandbox to allow bash.
Failed to exec /bin/bash as variant for /bin/sh (1: Operation not permitted).
This is used to determine the dependency tree of impure libraries so nix
knows what paths to open in the sandbox. With the less restrictive
defaults it isn't needed anymore.
Running `nix-store --gc --delete` will, as of Nix 2.3.3, simply fail
because the --delete option conflicts with the --delete operation.
$ nix-store --gc --delete
error: only one operation may be specified
Try 'nix-store --help' for more information.
Furthermore, it has been broken since at least Nix 0.16 (which was
released sometime in 2010), which means that any scripts which depend
on it should have been broken at least nine years ago. This commit
simply formally removes the option. There should be no actual difference
in behaviour as far as the user is concerned: it errors with the exact
same error message. The manual has been edited to remove any references
to the (now gone) --delete option.
Other information:
* Path for Nix 0.16 used:
/nix/store/rp3sgmskn0p0pj1ia2qwd5al6f6pinz4-nix-0.16
In
nix-instantiate --dry-run '<nixpkgs/nixos/release-combined.nix>' -A nixos.tests.simple.x86_64-linux
this reduces time spent in unparse() from 9.15% to 4.31%. The main
culprit was appending characters one at a time to the destination
string. Even though the string has enough capacity, push_back() still
needs to check this on every call.
The problem fixed: each nix-shell invocation creates a new temporary
directory (`/tmp/nix-shell-*`) and never cleans up.
And while I'm here, shellescape all variables inlined into the rcfile.
See what might happen without escaping:
$ export TZ="';echo pwned'"
$ nix-shell -p hello --run hello
pwned
Hello, world!
The ssh client is lazily started by the first worker thread, that
requires a ssh connection. To avoid the ssh client to be killed, when
the worker process is stopped, do not set PR_SET_PDEATHSIG.
