Without this, a `#include <float.h>` resolves incorrectly. Either the
headers weren't on the include path at all, or they only were for
local, not system, imports.
What's weird is this used to not be a problem. Not sure what other
change in e.g. cc-wrapper would affect this.
I think it's ok to export things which aren't wrapped. The cc-wrapper
can be thought of as responsible for all of binutils and the c
compiler, only wrapping those binaries which are necessary to
interposition---as opposed to all binaries it thinks are relevaant.
Conversely, adding the setup hook to the unwrapped compilers would be
unforunate as hooks are ugly hacks and the compilers themselves take
a long time to rebuild. Better to wholely separate "pure packages" from
hacks.
Packages get --host and --target by default, but can explicitly request
any subset to be passed as needed. See docs for more info.
rustc: Avoid hash breakage by using the old (ignored)
dontSetConfigureCross when not cross building
Eventually we should avoid this "pre-wrapping" and just update those
files in nixpkgs. This mass-rebuild change is best done along with
those needed to reduce the disparity between native and cross (i.e.
making native the "identity cross").
We now (on cross) require per-target flag interposition by putting the
triple in the names of the relevant environment variables, e.g:
export NIX_arm_unknown_linux_gnu_CFLAGS_COMPILE=...
The wrapper also has a `infixSalt` attribute (and "_" prefixed and
suffixed variants) to assist downstream packages.
Note how that the dashes are replaced to keep the identifier valid.
Using names like this allows us to keep the settings for different
compilers seperate.
I think it might be even better to use names like `NIX_{BUILD,HOST}...`
using the platform's role rather than the platform itself, but this
would be more work as the previous stages' tools would have to be re-
wrapped to take on their new role. I therefore didn't do this for now,
but that route should be thoroughly explored in the future.
Since 9c57f3b5c0 bumped the protobuf
version because the new upstream requires it, electrum now gets
protobuf3_0 *and* protobuf3_2 instead of just one version.
This leads to the following build errer:
Found duplicated packages in closure for dependency 'protobuf':
protobuf 3.0.2 (...-python2.7-protobuf-3.0.2/lib/python2.7/site-packages)
protobuf 3.2.0 (...-python2.7-protobuf-3.2.0/lib/python2.7/site-packages)
Using protobuf3_2 for keepkey and electrum fixes the build.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Cc: @np
Regression introduced by 76beb08313.
With version 0.7.15 a few additional dependencies are needed by trezor,
mainly a newer version of protobuf bindings and requests.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Cc: @np
The motivation is to be able to get rid of common configuration
when initial packages differs since common configuration assumes
a very specific version set.
cc @jmitchell @peti
With newer Nix it's (fortunately) no longer possible to create a file
with setuid bits, even though the permissions are fixed later the build
will fail during installPhase already.
I've verified whether the contents of the output path are the same as
before this change and the contents match.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Upstream changes:
* Added OpenSSL.X509Store.set_time() to set a custom verification time
when verifying certificate chains. pyca/pyopenssl#567
* Added a collection of functions for working with OCSP stapling. None
of these functions make it possible to validate OCSP assertions, only
to staple them into the handshake and to retrieve the stapled
assertion if provided. Users will need to write their own code to
handle OCSP assertions. We specifically added:
Context.set_ocsp_server_callback, Context.set_ocsp_client_callback,
and Connection.request_ocsp. pyca/pyopenssl#580
* Changed the SSL module's memory allocation policy to avoid zeroing
memory it allocates when unnecessary. This reduces CPU usage and
memory allocation time by an amount proportional to the size of the
allocation. For applications that process a lot of TLS data or that
use very lage allocations this can provide considerable performance
improvements. pyca/pyopenssl#578
* Automatically set SSL_CTX_set_ecdh_auto() on OpenSSL.SSL.Context.
pyca/pyopenssl#575
* Fix empty exceptions from OpenSSL.crypto.load_privatekey().
pyca/pyopenssl#581
The full upstream changelog can be found at:
https://pyopenssl.readthedocs.io/en/17.0.0/changelog.html
I've also added a patch from pyca/pyopenssl#637 in order to fix the
tests, which was the main reason for the version bump because that patch
won't apply for 16.2.0.
According to the upstream changelog there should be no
backwards-incompatible changes, but I've tested building against some of
the packages depending on pyopenssl anyway. Regardless of this, the
build for pyopenssl fails right now anyway, so the worst that could
happen via this commit would be that we break something that's already
broken.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Escape things by default in derivation names (i.e. digit cannot be the
first character etc.)
Update Quicklisp (tracking upstream); list new missing dependencies
Add some minimal README about ql-to-nix
> http://opus-codec.org/release/stable/2017/06/20/libopus-1_2.html
Changes since 1.1.x include:
- Speech quality improvements especially in the 12-20 kbit/s range
- Improved VBR encoding for hybrid mode
- More aggressive use of wider speech bandwidth, including fullband speech starting at 14 kbit/s
- Music quality improvements in the 32-48 kb/s range
- Generic and SSE CELT optimizations
- Support for directly encoding packets up to 120 ms
- DTX support for CELT mode
- SILK CBR improvements
- Support for all of the fixes in draft-ietf-codec-opus-update-06 (the mono downmix and the folding fixes need --enable-update-draft)
- Many bug fixes, including integer wrap-arounds discovered through fuzzing (no security implications)
Includes a more recent version of antlr to nixpkgs. Previous
versions exist already, but version 4 brings many changes
to the generated code and runtime targets.
The install location has been changed from previous versions
of antlr to make use of the set-java-classpath hook, which
is required to make use of both the runtime and the binary.
Also includes the testing rig as a script to allow graphical
inspection of parse trees.
llvm-config is a tool to output compile and linker flags, when compiling against llvm.
The tool however outputs static library names despite libllvm is build
as shared library on nixos. This was fixed for llvm 3.4, 3.5 and 3.7.
For llvm 3.8 and 3.9 it printed the library extension twice (.so.so).
This was fixed in 4.0 and the patch is backported to 3.8 and 3.9 in
this pull request.
```
$ for i in 34 35 37 38 39; do echo "\nllvm-$i"; nix-shell -p llvmPackages_$i.llvm --run 'llvm-config --libnames'; done
llvm-34
libLLVMInstrumentation.so libLLVMIRReader.so libLLVMAsmParser.so
...
llvm-35
libLLVMLTO.so libLLVMObjCARCOpts.so libLLVMLinker.so libLLVMipo.so
...
llvm-37
libLLVMLTO.so libLLVMObjCARCOpts.so libLLVMLinker.so libLLVMBitWriter.so
...
llvm-38
libLLVM-3.8.1.so
llvm-39
libLLVM-3.9.so
```
fixes#26713
The documentation got a bit stale compared to actual contents of
nixpkgs. This commit focuses on updating existing docs, not on making
sure all details of beam packages are covered.
This makes beam.package.erlangR19.abnf to be actually built with R19,
instead of the default R18. It means that Elixir and LFE are provided in
two versions, one built with R18 and with R19.
Please note that Elixir does not build with R16 and R17 - trying to
access beam.packages.erlang{R16,R17}.elixir will throw an error.
This change introduces a fixpoint, which allows to do deep override when
building packages defined in pkgs/development/beam-modules/default.hex.
This allows to provide beam.packages.erlang{,R16,R17,R18,R19} which
contains the same packages built with different Erlang/OTP versions.
Top-level attribute beamPackages points at beam.packages.erlangR18, the
same applies to other top-level Erlang packages.
TODO:
- beam.packages.erlang{R16,R17} is almost useless, since rebar/rebar3
does not build using these versions;
- all packages in beam.packages which use buildMix are actually built
with erlangR18;
- update documentation.
