The idris2 Makefile tries to use different versions of sed depending on
the OS, but nix always uses the same version. Because the version of
sed that is expected on macOS doesn't exist in the nix environment, the
build fails. Setting the OS to empty string resolves the issue.
rls has racer baked in which needs to know where the rust source
is to be able to do completion for std libs. By default rls will use:
$(rustc --print sysroot)/lib/rustlib/src/rust/src
which is nonexistent, this commit sets the correct source path
in a same way like it's done in racer expression.
reasoning:
sjlj (short jump long jump) exception handling makes no sense on x86_64, it's forcably slowing programs down as it produces a constant overhead. On x86_64 we have SEH (Structured Exception Handling) and we should use that. On i686, we do not have SEH, and have to use sjlj with dwarf2. Hence it's now conditional on x86_32
Programs which generate and compile a lot of code at runtime (such as
programming language interpreters like ACL2) are not suited for running on SBCL
executables built with the "immobile space" feature, as explained by Douglas
Katzman in this mail thread:
https://sourceforge.net/p/sbcl/mailman/message/36007057/
In this commit, I add an optional flag to the SBCL package allowing you to
disable the "immobile space" features.
I also migrated away from specifying enabled/disabled features in a
`customize-target-features.lisp` file and towards supplying them as command line
arguments to `make.sh`, as has been recommended by the installation instructions
since 2012 or so.
`libstdc++` and a few other libraries are comiled with the options
set in `EXTRA_TARGET_FLAGS`. Normally, this is filled form
`EXTRA_FLAGS` inside of `builder.sh`, from which it inherits its
optimization option. For cross compilers `EXTRA_TARGET_FLAGS` is
set by a dedicated function that does not specify any optimization,
leading to sub-par runtime performance of many C++ programs.
Darwin has a bug which affects the use of poll() with a tty fd,
which affects gambit's REPL when at a console, causing 100% CPU
usage.
Gambit recommends this is disabled on Darwin.
- Use the new Gambit support.
- Move files from $out to $out/gerbil.
- Use new Gerbil configuration and installation scripts.
- Move some fixups from preBuild to postPatch.
- Give up on previous failed attempts at using static libraries.
- Add support for compiling libraries written in Gerbil.
- Build using NIX_BUILD_CORES.
- Register all those things in all-packages.
Refactor the build rule:
- Put files in $out/gambit instead of $out.
- Make the optimization setting easy to override.
- Make use of gccStdenv more explicit at this level.
- Support new-style runtime options for forcing UTF-8 I/O.
- Override the PACKAGE_VERSION and PACKAGE_STRING with git version.
- Note that the license is lgpl21, not lpgl2 (Note: also dual asl20).
- Try and fail to meaningfully add missing runtimeDeps.
- Build using NIX_BUILD_CORES.
The compiler does not need it anymore, has not needed it for many years
iirc. This just goes in and pollutes the environment overriding the
users GOPATH and causing grief.
Go even warns about it itself, without vs with this commit:
```sh
~> go env GOPATH
/home/manny/go
~> nix-shell -p go
~> go env GOPATH
warning: GOPATH set to GOROOT (/nix/store/gvw1mfpdrk7i82884yhxf9lf5j3c12zm-go-1.14.1/share/go) has no effect
/nix/store/gvw1mfpdrk7i82884yhxf9lf5j3c12zm-go-1.14.1/share/go
~> exit
~> nix-shell -I nixpkgs=cloned/NixOS/nixpkgs -p go
~> go env GOPATH
/home/manny/go
~> exit
```
There are several tarballs (such as the `rust-lang/rust`-source) with a
`Cargo.toml` at root and several sub-packages (with their own Cargo.toml)
without using workspaces[1].
In such a case it's needed to move into a subdir to only build the
specified sub-package (e.g. `rustfmt` or `rsl`), however the artifacts
are at `/target` in the root-dir of the build environment. This breaks
the build since `buildRustPackage` searches for executables in `target`
(which is at the build-env's root) at the end of the `buildPhase`.
With the optional `buildAndTestSubdir`-argument, the builder moves into
the specified subdir using `pushd`/`popd` during `buildPhase` and
`checkPhase`.
Also moved the logic to find executables and libs to the end of the `buildPhase`
from a custom `postBuild`-hook to fix packages with custom `build`/`install`-procedures
such as `uutils-coreutils`.
[1] https://doc.rust-lang.org/book/ch14-03-cargo-workspaces.html
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-kenny did a good job in the past and is set as maintainer in many package,
however since 2017-2018 he stopped contributing. To create less confusion
in pull requests when people try to request his feedback, I removed him as
maintainer from all packages.
Kept 1.42 around for Thunderbird.
i686-apple-darwin is no longer supported upstream. We could still
support building it, for this one release, since we have the binary
for the previous release, (or bootstrap it for future releases from
Rust 1.42,) but since this release is the one that drops support, I
think it makes sense to do it now. (And probably nobody is using it
anyway.)
We've already worked around failing stat() calls in the installer by
running it in qemu-user. Turns out on some systems this workaround is
also needed to run `wlib', the archiver of the Open Watcom toolchain.
This issue manifested itself in broken VirtualBox builds due to
/build/virtualbox/out/linux.amd64/dbgopt/obj/VBoxPcBios32/pci32.obj
not being found by `wlib'.
We now just wrap all binaries in qemu-user to avoid this.
Everything is copied as-is from 9 (except version and hash).
Some platform-specific patches might not apply anymore;
I'm lazily leaving that for the community to fix.
The nss update is needed for security update of firefox.
For linux platforms only about 1k aarch64 rebuilds are missing;
the diff on Hydra looks OK. Darwin needs 20k more rebuilds,
but I don't think we want to wait for that.
This option can be used to set the “jit” language which enable the
libgccjit functionality. Also adds a “libgccjit” attr which is gcc
built with just jit enabled.
libgccjit is a library but is used as a compiler. So it references a
bunch of compiler things in $out. To avoid a cycle, we need to put
everything in $out, so referenced to $lib need to be replaced with
${!outputLib}.
Some elm packages (eg. https://github.com/NoRedInk/elm-simple-fuzzy/ ) have a Makefile which the standard builder tries to build and fails because of missing dependencies. This change forces a no-op configure and build phase to avoid such a situation.
* treewide Drop unneeded go 1.12 overrides
* Fix packr to be go module compatible.
I updated to version 2.8.0 which is the latest on master.
Then due to the 2 different sets of go modules which are used, I split
the build into two different derivations, then merged them togethor
using symlinkJoin to have the same output structure as the existing derivation.
* Remove consul dependency on go1.12
I updated the consul version to 1.7.2 and flipped it to building using
modules.
* Remove go1.12 from perkeep.
Update the version to the latest unstable on master.
* Update scaleway-cli to not be pinned to go1.12
Switched the version to 1.20
* Update prometheus-varnish-exporter to not depend on go1.12
* Update lnd to build with go1.12
Updated the version
Forced only building subpackages with main to prevent panics over
multiple modules in one repo
* Remove go1.12 from openshift
Had to update the version to 4.1.0 and do a bit of munging to get this
to work
* Remove go1.12 completely.
These are no longer needed.
* Update bazel-watcher and make it build with go 1.14
`jellyfin` appeared unsupported on `aarch64` due to `dotnet` platform
support in nixpkgs, but there are ARM64 downloads of the `dotnet` SDKs
available. This change follows the kind of pattern used in the
`firecracker` packaging to support selective x86_64/arm64 downloads.
With this change I can build `jellyfin` on a Raspberry Pi 4. The other
content hashes have been filled in, and all build successfully, but
they have not been further tested.
This includes several enhancements in the underlying compiler, including
codegen improvements for AVX-512, Ice Lake CPU definitions,
cross-{arch,os} compilation (currently unsupported due to multilib
issues), and more.
This also bumps the LLVM backend to the 10.0 release. Note that ispc
itself requires a few extra stability patches on top of 10.0 for AVX-512
support, but these aren't applied for us. Therefore AVX-512 still has
some extra, rough edges.
Signed-off-by: Austin Seipp <aseipp@pobox.com>
This is a better name since we have multiple 64-bit things that could
be referred to.
LP64 : integer=32, long=64, pointer=64
ILP64 : integer=64, long=64, pointer=64
In order to use OVMF firmware with e.g. qemu on macOS, these packages
needed to be made macOS ready. This meant choosing the clang build in
this case, because it is the only one working on macOS.
Unfortunately, just using clang on all platforms doesn't work because
there are hardcoded assumptions in the edk2 build system.
This makes packages use lapack and blas, which can wrap different
BLAS/LAPACK implementations.
treewide: cleanup from blas/lapack changes
A few issues in the original treewide:
- can’t assume blas64 is a bool
- unused commented code
* ghcHEAD: bump to 8.11.20200403
* ghcHead: reduce diff vs. 8.10.1
dontAddExtraLibs was removed by accident (IMO) in ea19a8ed1e
* ghcHEAD: add ability to use system libffi
- enable nixpkgs' libffi
- minimise diffs against 8.10.1
- remove patching
* remove configure warning about --with-curses-includes
configure: WARNING: unrecognized options: --with-curses-includes
