The `platform` field is pointless nesting: it's just stuff that happens
to be defined together, and that should be an implementation detail.
This instead makes `linux-kernel` and `gcc` top level fields in platform
configs. They join `rustc` there [all are optional], which was put there
and not in `platform` in anticipation of a change like this.
`linux-kernel.arch` in particular also becomes `linuxArch`, to match the
other `*Arch`es.
The next step after is this to combine the *specific* machines from
`lib.systems.platforms` with `lib.systems.examples`, keeping just the
"multiplatform" ones for defaulting.
By exporting it, we always make the new directories available
to subprocesses, regardless of whether the environment
variable existed before `nix-shell` was invoked.
This avoids the scenario where strictDeps is off and cross-compiled
XDG_DATA_DIRS content is brought into the environment.
While probably harmless for data like manpages and completion scripts,
this would cause issues when XDG_DATA_DIRS is used to find executables
or plugins. The Qt framework is known to behave like this and might
have run into incompatibilities.
XDG_DATA_DIRS is to /share as PATH is to /bin.
It was defined as part of the XDG basedir specification.
https://specifications.freedesktop.org/basedir-spec/basedir-spec-latest.html
While it originated from the X Desktop Group, it is not limited to
the X11 ecosystem, as evidenced by its use in bash-completion.
The removal of ` && -d "$pkg/bin"` is ok, because this optimization is
already performed by `addToSearchPath`.
This adds -frandom-seed to each compiler invocation in stdenv. The
object here is to make the compierl invocations produce the same output
every time they are called (for the same derivation). When the
-frandom-seed option is not set the compiler will use a combination of
random numbers (in GCC's case from /dev/urandom) and the durrent time to
produce a "random" input per file. This can (among other things) lead to
different ordering of symbols in the produced object files.
For reason of reproducibility we prefer having the same derivation
produce the exact same outputs. This is not a silver bullet but one way
to tame the compiler.
defaultHardeningFlags is set to enable pie for Musl, but is not
actually used because the default is never put into
NIX_HARDENING_ENABLE. That still works for cases other than Musl
only because NIX_HARDENING_ENABLE is defaulted in the binutils and
cc-wrapper setup-hook.sh scripts.
This hook moves systemd user service file from `lib/systemd/user` to
`share/systemd/user`. This is to allow systemd to find the user
services when installed into a user profile. The `lib/systemd/user`
path does not work since `lib` is not in `XDG_DATA_DIRS`.
This introduces the .inputDerivation attribute on all derivations
created with mkDerivation. This is another derivation that can always
build successfully and whose runtime dependencies are the build time
dependencies of the original derivation.
This allows easy building and distributing of all derivations needed to
enter a nix-shell with
nix-build shell.nix -A inputDerivation
I hate the thing too even though I made it, and rather just get rid of
it. But we can't do that yet. In the meantime, this brings us more
inline with autoconf and will make it slightly easier for me to write a
pkg-config wrapper, which we need.
The cross file is added in the `mkDerivation`. It isn't nice putting
build tool-specific stuff here, but our current architecture gives us
little alternative.
Currently it's not possible to determine the reason why a package is
unavailable without evaluating nixpkgs multiple times with different
settings. eg.
nix-repl> :p android-studio.meta
{ available = false; broken = false; unfree = true; unsupported = true; ... }
The following snippet is an example that uses this information to query
the availability information of all packages in nixpkgs, giving an
overview of all the packages currently marked as broken, etc.
{ pkgs }:
with import <nixpkgs/lib>;
let
mapPkgs =
let mapPkgs' = path: f: mapAttrs (n: v:
let result = builtins.tryEval (v ? meta); in
if !result.success then {} else
if isDerivation v then f (path ++ [n]) v else
if isAttrs v && v.recurseForDerivations or false then mapPkgs' (path ++ [n]) f v else
{}
);
in mapPkgs' [];
getMeta = path: drv:
if drv.meta ? available then
let meta = {
pkg = concatStringsSep "." path;
inherit (drv.meta) broken unfree unsupported insecure;
};
in builtins.trace meta.pkg meta
else {};
metaToList = attrs: flatten (map (v: if v ? pkg then v else metaToList v) (attrValues attrs));
in metaToList (mapPkgs getMeta pkgs)
While looking at the graph of all the outputs in my personal binary
cache it became obvious that we have a lot of self references within the
package set. That isn't an isuse by itself. However it increases the
size of the binary cache for every (reproducible) build of a package
that carries references to itself. You can no longer deduplicate the
outputs since they are all unique. One of the ways to get rid of (a few)
references is to rewrite all the symlinks that are currently used to be
relative symlinks. Two build of something that didn't really change but
carries a self-reference can the be store as the same NAR file again.
I quickly hacked together this change to see if that would yield and
success. My bash scripting skills are probably not great but so far it
seem to somewhat work.
Before, we'd always use `cc = null`, and check for that. The problem is
this breaks for cross compilation to platforms that don't support a C
compiler.
It's a very subtle issue. One might think there is no problem because we
have `stdenvNoCC`, and presumably one would only build derivations that
use that. The problem is that one still wants to use tools at build-time
that are themselves built with a C compiler, and those are gotten via
"splicing". The runtime version of those deps will explode, but the
build time / `buildPackages` versions of those deps will be fine, and
splicing attempts to work this by using `builtins.tryEval` to filter out
any broken "higher priority" packages (runtime is the default and
highest priority) so that both `foo` and `foo.nativeDrv` works.
However, `tryEval` only catches certain evaluation failures (e.g.
exceptions), and not arbitrary failures (such as `cc.attr` when `cc` is
null). This means `tryEval` fails to let us use our build time deps, and
everything comes apart.
The right solution is, as usually, to get rid of splicing. Or, baring
that, to make it so `foo` never works and one has to explicitly do
`foo.*`. But that is a much larger change, and certaily one unsuitable
to be backported to stable.
Given that, we instead make an exception-throwing `cc` attribute, and
create a `hasCC` attribute for those derivations which wish to
condtionally use a C compiler: instead of doing `stdenv.cc or null ==
null` or something similar, one does `stdenv.hasCC`. This allows quering
without "tripping" the exception, while also allowing `tryEval` to work.
No platform without a C compiler is yet wired up by default. That will
be done in a following commit.
Rewrite the `stripHash` helper function with 2 differences:
* Paths starting with `--` will no longer produce an error.
* Use Bash string manipulation instead of shelling out to `grep` and
`cut`. This should be faster.
A bunch of stdenv-internal variables were deleted in
1601a7fcce, but these are needed in the
fixup phase, whereas the rest are just needed for the initial work
(findInputs, etc) before the user phases.
CC @matthewbauer
