The old boot.spl.hostid option was not working correctly due to an
upstream bug.
Instead, now we will create the /etc/hostid file so that all applications
(including the ZFS kernel modules, ZFS user-space applications and other
unrelated programs) pick-up the same system-wide host id. Note that glibc
(and by extension, the `hostid` program) also respect the host id configured in
/etc/hostid, if it exists.
The hostid option is now mandatory when using ZFS because otherwise, ZFS will
require you to force-import your ZFS pools if you want to use them, which is
undesirable because it disables some of the checks that ZFS does to make sure it
is safe to import a ZFS pool.
The /etc/hostid file must also exist when booting the initrd, before the SPL
kernel module is loaded, so that ZFS picks up the hostid correctly.
The complexity in creating the /etc/hostid file is due to having to
write the host ID as a 32-bit binary value, taking into account the
endianness of the machine, while using only shell commands and/or simple
utilities (to avoid exploding the size of the initrd).
Getting the names of all environment variables is tricky. The previous
implementation easily got confused by multi-line variables. The new
one is more reliable but not still not perfect.
This works around a segfault in Bash 4.3, where the expression
"${!var}" (where var="-9") crashes under certain conditions.
http://hydra.nixos.org/build/16693445
Otherwise, stdenv won't have a reference to e.g. patchelf on Linux
(because it was passed in by mkDerivation). This causes the installer
tests to fail, because having "stdenv" in the installation CD closure
is not enough to pull in all stdenv packages.
http://hydra.nixos.org/build/16546643
This allows licenses like the Amazon Software License to be identified
properly while still preventing packages with those licenses from
being distributed in the Nixpkgs/NixOS channels.
This makes stuff build with Xcode 6.1 on Mac OS X 10.9 (where we got
errors like "ld: file not found:
/usr/lib/system/libsystem_coreservices.dylib for architecture x86_64" due to the use of the 10.10 SDK).
The static curl program is gone, replaced by curl inside of the
bootstrap tools tarball. Also, we generate a .tar.xz archive rather
than .cpio.bz2, making the download smaller. The separate
{sh,cpio,mkdir,ln,bzip2} programs have been replaced by a single
busybox program.
In 3.3, a C++ class defined in a header will get a typeinfo symbol
like this (e.g. in Nix's src/libutil/util.o):
(__DATA,__datacoal_nt) weak external typeinfo for nix::BaseError
But in 3.4, this has changed to:
(__DATA,__datacoal_nt) weak external automatically hidden typeinfo for nix::BaseError
This causes the linker to change the symbol to:
(__DATA,__data) non-external (was signed char private external) typeinfo for nix::BaseError
i.e. losing its weak linkage. But without weak linkage, dynamic_cast
and other RTTI-based mechanisms (such as catching an exception of a
certain type) don't work across shared libraries / executables.
The clang compiler in the SDK doesn't have this behaviour, but it's
not clear exactly which version it is (it just says "based on LLVM
3.4svn").
This should fix the OpenJDK build, which was failing because paxctl is
in sbin and therefore not automatically added to $PATH.
http://hydra.nixos.org/build/15658346
Copying /usr/lib/system/libunwind.dylib at evaluation time doesn't
work (e.g. on Hydra). And copying binary system libraries is a bad
idea anyway for license reasons.
Make thread disabling explicit. This changes the semantics of the perl
derivation, so on other platforms it may require setting
enableThreading = false
This commit doesn't change the derivation or out hash of stdenvLinux.
The stage3.extraAttrs.glibc argument was required for this whole build
procedure to correctly work and it was very-very hard to see why (the
comment said something about gcc47, but we're using gcc48 now).
This stage3.extraAttrs.glibc goes into stage3.stdenv.glibc after some
arg passing, and in pkgs/development/compiler/gcc an (stdenv ? glibc)
boolean expression decides to override /usr/include during the GCC
build.
All of our stages are built with glibc, so this refactoring moves this
repeating specification of glibc (once for gcc wrapper and once here for
extraAttrs) to stageFun, by getting rid of wrapGCC, as we were using
that in all of the stages anyways.
Incidentally it turned out, that this stdenv.glibc inconsistency caused
some random other stuff to behave differently:
- stage1.pkgs.perl has threading disabled,
- stage4.pkgs.coreutils (the production coreutils) has testing disabled.
Leave this historical accidents as they are in this commit, so the scope
of this commit can stay as a refactoring only, these issues will be
fixed in separate commits.
This commit doesn't change the derivation or the output hash of
stdenvLinux.
Use the new allowedRequisites feature in stdenvLinux.
This way we properly check that the end-result stdenv of the quite
complicated multi-stage stdenvLinux building procedure is sane, and only
depends on the stuff that we know about.
Alternative would be to just disallowRequisites bootstrapTools, which is
the most common offender, but we have had other offenders in the past.
For these checks to actually fire, you currently have to use nixUnstable,
as the necessary feature will be released in Nix 1.8.
Previously stdenv depended on two different zlibs and there was a third
one in the top-level package set for other purposes. This commit merges
all this zlibs to one.
Actually this have been committed once as 1f2b636, but then got lost
while resolving merge conflicts. Hopefully it survives this time.
The point here is that it's always possible to debug this staging
logic by using `nix-repl pkgs/stdenv/linux'. The modular-stdenv
change introduced the lib dependency, which we now default to
../../../lib.
No derivation or out hashes of stdenvLinux is changed by this commit.
Previously stdenv depended on two different zlibs and there was a third
one in the top-level package set for other purposes. This commit merges
all this zlibs to one.
Make stages explicit and generalize the pattern of having an stdenv and
a pkgs collection for all stages to a common stage generating function
called stageFun.
Rewrite all stage handling with this new function.
This commit doesn't change the outhash (or drvhash) of the stdenv.
Don't use default parameter values, to make the callsites more readable
and for easier debuggability/changability. Also reordered the
callsites' parameter ordering for consistency.
In the final stdenv don't repeat the name of the shell.
This commit doesn't change the outhash (or drvhash) of the stdenv.
All the different stages of stdenv had the fetchurl inherited anyways,
so make this generic in stdenvBootFun.
This commit doesn't change the outhash (or drvhash) of the stdenv.
Now gcc is just another build input, making it possible in the future
to have a stdenv that doesn't depend on a C compiler. This is very
useful on NixOS, since it would allow trivial builders like
writeTextFile to work without pulling in the C compiler.
If $src refers to a directory, then always copy it. Previously, we
checked the extension first, so if the directory had an extension like
.tar, unpackPhase would fail.
If a build input is a regular file, use it as a setup hook. This makes
setup hooks more efficient to create: you don't need a derivation that
copies them to $out/nix-support/setup-hook, instead you can use the
file as is.
You can now register multiple values per named hook, e.g.
addHook preConfigure "echo foo"
addHook preConfigure "echo bar"
will cause ‘runHook preConfigure’ to run both ‘echo foo’ and ‘echo
bar’ (in that order). It will also call the shell function
preConfigure() or eval the shell variable $preConfigure, if
defined. Thus, if you don't call addHook, it works like the old hook
mechanism.
Allowing multiple hooks makes stdenv more modular and extensible. For
instance, multiple setup hooks can define a preFixup hook, and all of
these will be executed.
This variable sets the minimal Mac OS X version required for
running binaries produced by the Darwin toolchain. Since it
defaults to the version of the user's SDK, setting it explicitly
should make our builds more deterministic. It's now set to 10.6
because that's what hydra.nixos.org runs.
Commit 262c21ed46 purported to enable
ignoreNulls, but it was bogus because it set the flag on the wrong
derivation (i.e. stdenv rather than the result of mkDerivation).
Recent versions of Xcode don't install headers in /usr/include but
in a directory like
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk/usr/include
So use that instead, falling back to /usr/include in case of an older
version of Xcode.
Commit 986f361946 started to use
<nix/fetchurl.nix> to "download" the bootstrap binaries from the
Nixpkgs tree, using the file:/// scheme. This has really bad
consequences:
* It makes any derivation depend on the path of the Nixpkgs tree. So
evaluating a package will produce a different .drv file when run
from different locations. No wonder Hydra evaluation has been so
slow lately: for every Nixpkgs evaluation, it had to create tens of
thousands of .drv files, even if nothing had changed.
* It requires the builder to have file system access to the Nixpkgs
tree. So if your tree is in your home directory, the stdenv
bootstrap would probably fail.
So now the binaries are downloaded from tarballs.nixos.org.
Also dropped PowerPC "support".
Stdenv adapters are kinda weird and un-idiomatic (especially when they
don't actually change stdenv). It's more idiomatic to say
buildInputs = [ makeCoverageAnalysisReport ];
This removes the need for hacks like stdenv.regenerate. It also
ensures that overrideGCC is now stackable (so ‘stdenv = useGoldLinker
clangStdenv’ works).
