With meson 0.47.0 (or 0.47.1, or git)
things are very wrong re:rpath handling
resulting in at best missing libs but
even corrupt binaries :(.
When we run patchelf it masks the problem
by removing obviously busted paths.
Which is probably why this wasn't noticed immediately.
Unfortunately the binary already
has a long series of paths scribbled
in a space intended for a much smaller string;
in my testing it was something like
lengths were 67 with 300+ written to it.
I think we've reported the relevant issues upstream,
but unfortunately it appears our patches
are what introduces the overwrite/corruption
(by no longer being correct in what they assume)
This doesn't look so bad to fix but it's
not something I can spend more time on
at the moment.
--
Interestingly the overwritten string data
(because it is scribbled past the bounds)
remains in the binary and is why we're suddenly
seeing unexpected references in various builds
-- notably this is is the reason we're
seeing the "extra-utils" breakage
that entirely crippled NixOS on master
(and probably on staging before?).
Fixes#43650.
This reverts commit 305ac4dade.
(cherry picked from commit 273d68eff8f7b6cd4ebed3718e5078a0f43cb55d)
Signed-off-by: Domen Kožar <domen@dev.si>
They are both as powerful, but buildEnv is treacherous: if you pass a
package which depends on another python (for example the one of unstable
when you are on stable) it will be *silently* dropped, leading to hair
pulling.
Use case:
override neovim from unstable, but still keep stable's pythonPackages.
This solves the following bug:
opening neovim in nix-shell -p pythonPackages.numpy does not enable to
run successfully
:!python -c "import numpy"
because the PYTHONPATH is wiped by the neovim wrapper.
This wiping is necessary for the python providers, though, otherwise a
python2 nix-shell will make the python3 provider read python2 files.
We wrap the providers only, instead of neovim as whole.
In 3027bca, binutils was upgraded from 2.28.1 to 2.30. However, in 2.30,
the ldmain.c file within binutils, which the nixpkgs new-dtags.patch
file is meant to modify, was changed in such a way that the patch no
longer works. As a result, the new dtags are not actually enabled, and
binaries are built with RPATH set instead of RUNPATH, thereby preventing
LD_LIBRARY_PATH from overriding this built-in path. This change corrects
this. The patch file is no longer necessary because binutils's ldmain.c
now sets link_info.new_dtags based on the configuration flags.
This was probably not noticed immediately because, when the derivation
is built with nix-build, the fixupPhase runs patchelf --shrink-rpath.
patchelf converts any RPATH in the binary into RUNPATH (unless
--force-rpath is specified). Of course, if the binary is built without
nix-build (such as in a nix-shell), this never occurs, and any RPATH in
the binary is left in place.
This changes the way how the `glib` dependecy cycles are broken to make
the tested `glib` the default.
`glib-tested` is no longer needed as packages now use the checked version when
`doCheckByDefault` is set, since it was not used anywhere, I removed it.
`depsHostBuild` is not a thing, would never be a thing per the rules,
and isn't used anywhere. This is just my typo, hitherto unnoticed
because "host -> host" dependencies are by far the most obscure form.
Domen Kožar