Before this commit, cc-wrapper/default.nix was using
`isGccArchSupported` to validate `-mtune=` values. This has two
problems:
- On x86, `-mtune=` can take the same values as `-march`, plus two
additional values `generic` and `intel` which are not valid for
`-march`.
- On ARM, `-mtune=` does not take the same values as `-march=`;
instead it takes the same values as `-mcpu`.
This commit fixes these two problems by adding a new
`isGccTuneSupported` function. For `isx86` this returns `true` for
the two special values and otherwise defers to `isGccArchSupported`.
This commit also adds support for `-mtune=` on Aarch64.
Unfortunately on Aarch64, Clang does not accept as wide a variety of
`-mtune=` values as Gcc does. In particular, Clang does not tune
for big.LITTLE mixed-model chips like the very popular RK3399, which
is targeted using `-march=cortex-a72.cortex-a53` in gcc.
To address this problem, this commit also adds a function
`findBestTuneApproximation` which can be used to map
clang-unsupported tunings like `cortex-a72.cortex-a53` to
less-precise tunings like `cortex-a53`.
The work which led to this commit arose because we now have
packages, like `crosvm`, which use *both* `clang` *and* `gcc`.
Previously I had been using `overrideAttrs` to set
`NIX_CFLAGS_COMPILE` on a package-by-package basis based on which
compiler that package used. Since we now have packages which use
*both* compilers, this strategy no longer works.
I briefly considered splitting `NIX_CFLAGS_COMPILE` into
`NIX_CFLAGS_COMPILE_GCC` and `NIX_CFLAGS_COMPILE_CLANG`, but since
`NIX_CFLAGS_COMPILE` is sort of a hack to begin with I figured that
adding the logic to `cc-wrapper` would be preferable.
* doc/stdenv: rewrite manual build procedure to be closer to an auto-build
This is based on
<https://jade.fyi/blog/building-nix-derivations-manually/> plus some
more original research.
The previous version of this section did not work for your choice of
simple Haskell package, e.g. haskellPackages.hscolour, due to things
like `compileBuildDriverPhase` and other custom phases that it
does not address at all.
It seems more correct to use genericBuild in development to harmonize it
with what is actually done.
I feel a little bit like I am committing a sin by suggesting using the
experimental CLI in the manual (afaict I am the first to do this), but I
have given the old version of the command, and there are justifiable
reasons to do it:
* The noted limitations with env-vars are fixed. The one with the
non-empty temp directory was one I ran into myself and oh boy was that
not fun to debug.
* Additionally the outputs are set *before* sourcing `setup.sh`: there
is an issue with nix-shell where the original version of `$out` winds
up in `NIX_LDFLAGS` due to _addRpathPrefix, which means that resulting
executables may not run properly.
It is sad that `nix develop` propagates a wrong value of `SHELL` to
builders, though. It is equally sad that `nix-shell` is essentially
abandoned upstream, with undocumented and not insignificant differences
from `nix develop`.
For the exact script differences:
https://github.com/NixOS/nix/blob/17e6b85d05b3d32df244b1d4e89aa41fd8bdcae8/src/nix-build/nix-build.cc#L516-L551https://github.com/NixOS/nix/blob/db026103b18fb8b5a719594502edd0f89eb9c268/src/nix/get-env.sh
Co-authored-by: Valentin Gagarin <valentin.gagarin@tweag.io>
Gaetan Lepage