This causes collisions between the build outputs of `nix` when building
in an environment with `nix1.perl-bindings` and `nix`:
```
collision between `/nix/store/aa4rrcj7dg2xj4rfkiclcmp745ibqng0-nix-2.0.4/lib/libnixstore.so' and `/nix/store/sp0sdi4bll80h58big1iy8kkh3qqxpw2-nix-1.11.16/lib/libnixstore.so'
builder for '/nix/store/wgbccin107lhm8cv9imnnvkx1j2pgibc-hydra-perl-deps.drv' failed with exit code 25
```
Unlike on linux these are not namespaced per user so this will cause
build failures if /tmp/nix-test was not removed by a previous build if
the nixbld user id doesn't match by accident. Nix already creates a
unique tempdir for builds so we can use that instead.
Fixes#44172
This makes the command ‘nix-env -qa -f. --arg config '{skipAliases =
true;}'’ work in Nixpkgs.
Misc...
- qtikz: use libsForQt5.callPackage
This ensures we get the right poppler.
- rewrites:
docbook5_xsl -> docbook_xsl_ns
docbook_xml_xslt -> docbook_xsl
diffpdf: fixup
The isSeccomputable flag treated Linux without seccomp as just a
normal variant, when it really should be treated as a special case
incurring complexity debt to support.
This reverts commit 695027f61c.
We really can't have "nix-env -i nix" *not* upgrade nixUnstable to a
newer nixStable. For instance, it would cause "nix upgrade-nix" to
produce a user environment with collisions.
as discussed in bed3695848
Different names make it easier for the users to notice updates to Nix stable,
and to have Nix stable and unstable simultaneously.
This reverts commit bed3695848.
This commit effectively makes it impossible to upgrade to nixUnstable
with nix-env without mucking about with priorities, as you can't
uninstall the old nix transactionally with the new nix and if you
uninstall the old one first you no longer have nix at your disposal to
install the new one.
This reverts commit 9711aac642.
Main change: glibc: 2.25-x -> 2.26-y, containing security fixes,
and various features and deprecations. Unfortunately, some of the
latter still cause (transitively) a couple hundred newly failing jobs.
I'm not delaying anymore, so that we have the security fix on master.
I mainly patched gcc, llvm and icu, but I can't fix everything...
