| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The "vim_configurable" derivation has had a long history in nixpkgs back
then when there was no RFC process and where people still were figuring
out better ways on how to configure compile-time flags.
One of those was a composableDerivation function, which was used for
"vim_configurable" and mapped attribute sets to autoconf-flags, so that
for example if there was a "--enable-foo" flag you could just use
something like:
vim_configurable.merge { cfg.fooSupport = true; }
You'd then get a Vim with "--enable-foo" passed to configureFlags.
However, the composableDerivation feature was too complicated and was
ultimately removed at some point. While it does allow for things such as
introducing new "edf" (stands for Enable Disable Feature and maps the
autoconf flags mentioned above to attribute sets) flags, the complexity
that comes with that system is way too large than using something like
eg.:
vim-full.overrideAttrs (drv: {
configureFlags = (drv.configureFlags or []) ++ [ "--enable-foo" ];
})
While this looks more verbose than the above, one can easily follow
what's happening, whereas if you'd need to add and enable a new "edf"
flag, you'd do something like this:
vim_configurable.merge {
flags = composableDerivation.edf { name = "foo"; };
cfg.fooSupport = true;
}
I admit that this does look a little nicer, but even I'm not sure
whether it's worth adding so much complexity since in practice I rarely
came across a sitation where something like the above would be really
beneficial.
So back then when "vim_configurable" was introduced[1], it was used as
an alternative to the main vim derivation but using composableDerivation
instead.
Nowadays however, vim_configurable no longer uses composableDerivation
and the rename also doesn't change any features, so I think it's safe to
rename vim_configurable to vim-full in Vuizvui.
[1]: https://github.com/NixOS/nixpkgs/commit/9a4e9e7a3b4014bb3c9f678ec22d254b85c4c98a
[2]: https://github.com/NixOS/nixpkgs/commit/4e5ebcc3ed1de9c5c2001c7d5829f4566e0bde3f
Signed-off-by: aszlig <aszlig@nix.build>
Cc: @Profpatsch
Cc: @devhell
|
|
|
|
|
|
|
|
|
| |
Since shakti's k9 is no longer available for download, this is quite
nonsensical to keep around. When it becomes available again, I'll be
better off downloading it and saving it locally somewhere. Nixified
shakti will just end up being garbage collected…
The k-gpp idea wasn't bad, but not great either.
|
|
|
|
|
|
| |
I'm particularly annoyed at the way `vim-markdown` handles folding, so
I'm going to give `markdown-preview-nvim` a spin. It's also actively
maintained and seems quite powerful.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The new Tridactyl native messenger is no longer written in Python, so
the path to the main executable is no longer valid.
I also trimmed down the allowed_extensions attribute to just include the
actual extension.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I had this laying around locally for a year now and I'm still not really
happy with some things, for example not having yet full source builds of
the extensions and a few config options not yet managed by Nix (eg.
search engines).
However, since Firefox takes a while to build it's a bit tedious to
always do it directly (and locally) after I update my machines. Having
this part of my workstation profile should make sure that my version of
Firefox is available at all times.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
| |
After https://github.com/NixOS/nixpkgs/pull/183008 it seems to be
necessary to set PLAN9 even with Sören's patch to resolve the font
acme needs. I don't have time to debug this issue at the moment,
so let's give in and set the variable.
|
|
|
|
| |
Fixes build (!) of my emacs on i686-linux.
|
|
|
|
| |
Fixes eval of my emacs on i686-linux
|
|
|
|
|
|
|
|
|
| |
depot reimports nixpkgs, so we need to make sure it passes the correct
value for `system`. As a result of how depot works, cross is unsupported
at the moment.
This breaks machines/sternenseemann/ludwig for the moment, which we'll
be able to address separately.
|
| |
|
|
|
|
|
| |
Disalbes checkMeta for depot which should prevent the odd eval failure
due to extra meta attrs it uses
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The isHydraChannel meta attribute is needed in order to tell Hydra that
the derivation in question should be a channel tarball. However in
Nixpkgs the meta attribute is not used, so checkMeta doesn't recognise
it as a valid attribute which leads to an evaluation error.
Recently[1] a commit got merged, which enables shallow type checking for
meta attributes by default. This led to an evaluation error for our
Hydra machine channels for the reason mentioned above.
I opted to work around that issue by adding meta.isHydraChannel after
mkDerivation, because adding isHydraChannel as a valid meta attribute to
Nixpkgs doesn't feel right to me since it's only relevant for Hydra and
its apparently deprecated[2] channel feature.
[1]: https://github.com/NixOS/nixpkgs/commit/6762de9a28e248f46bd0810e03c
[2]: https://github.com/NixOS/hydra/blob/53335323ae79ca1a42643f58e520b376898ce641/doc/manual/src/jobs.md#meta-fields
Signed-off-by: aszlig <aszlig@nix.build>
|
| |
|
|
|
|
| |
Contains adjustments for recent alias removals in nixpkgs.
|
| |
|
|
|
|
| |
The Qr-code was done with a strange format that only allowed ASCII.
|
|
|
|
|
|
|
|
|
|
|
| |
This attribute has been renamed a while ago[1] but recently even the
alias was removed[2], which now triggers an evaluation error in Vuizvui.
[1]: https://github.com/NixOS/nixpkgs/commit/bc49a0815ae860010b4d593b02f
[2]: https://github.com/NixOS/nixpkgs/commit/ba3319568df2c6675dbe36478fb
Signed-off-by: aszlig <aszlig@nix.build>
Cc: @Profpatsch
|
|
|
|
|
|
|
|
|
|
|
|
| |
Found this thanks to @fpletz starring the repository and since I
constantly use f/F and t/T when navigating, I decided to give it a try.
Usually when trying things, I'd just add the Vim version with the plugin
to my local Nix profile, but since I'm already irritated when other
things than the cursor are highlighted in the current line, I decided to
do something completely irrational and just add it to all my machines.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Some personal history on this:
I started to get used to AT&T syntax because it's the default in GDB and
used that ever since until I one day starting to do some more reverse
engineering using radare, which defaults to Intel syntax.
Ever since then probably my most used command in GDB was "set
disassembly-flavor intel" (because I was to lazy to add it to the
config) because I constantly got confused by the source/destination
operand swaps. This even happened during live reverse engineering at rC3
where I was confused about some function logic only to find out that I
was viewing in AT&T syntax.
Fast-forward to today: I'm debugging some application using WINE and
winedbg uses AT&T syntax, which I didn't like at first. After reflecting
on this for a while, I thought it would probably be better to get used
to AT&T syntax again and switch everything to use AT&T for the following
reasons:
* Operands are more natural to read, since most libraries/APIs in
higher level languages do it like this (well, except memcpy, strcpy,
etc... maybe I now get confused by libc functions...)
* AT&T syntax feels less verbose, for example "mov ecx, dword [eax]"
is just "movl (%eax), %ecx"
This very commit makes sure that radare2 now defaults to AT&T syntax
instead of eg. ensuring that GDB uses Intel syntax by default.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When using NIX_SANDBOX_DEBUG_INJECT_FILES (which we now call
NIX_SANDBOX_DEBUG_INJECT_DIRS, because it's more accurate), I usually
used it to provide fake /dev or /sys directories.
I turned out, that today I was trying to use this functionality again
(who'd have known) and it also turned out that I forgot to create the
target directory, which wasn't needed back then for /dev or /sys because
they were already existing.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
|
|
| |
With my new laptop, a font size of 12pt is rather large and given that
hidpi displays usually have a quite large resolution (the name might
hint at that), we don't necessarily need to use embedded bitmaps anymore
which was one of the reasons why I used a point size of 12.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
|
|
|
|
|
| |
We only want to highlight #aabbcc-style hex colors and not every damn
hex value that happens to be 3 or 4 bytes, which is very annoying when
you're doing a lot of bit manipulation.
Additionally, I currently have a code base where "gold" is something a
player can have, so one doesn't need much imagination on how irritating
that is.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
| |
The config we're baking in by default relies on Nightly features, so it
only makes sense to actually use a rustfmt version that supports it.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
| |
Seems like something got more strict about setup hook execution and the
relevant env vars for gi weren't set in checkPhase.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We no longer need to use extraPostFetch in order to append stuff to
the postFetch phase of fetchzip. Since this got fixed[1], we now get
appropriate warnings during evaluation:
warning: use 'postFetch' instead of 'extraPostFetch' with 'fetchzip'
and 'fetchFromGitHub'.
Neither do we like warnings nor do we like to use workarounds from the
past that got fixed, so let's fix it on our side as well :-)
[1]: https://github.com/NixOS/nixpkgs/pull/173430
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
|
|
| |
I don’t know how I missed this, but my deploy script didn’t actually
set up a system profile, so rebooting would lead to (very) old
generations being activated again.
Thanks to @sternenseemann for the nix-env trick (via
https://code.tvl.fyi/tree/ops/nixos.nix#n55 )
|
|
|
|
| |
idk why/how this works and tbh I don’t even want to know.
|
|
|
|
| |
thx to @sternenseemann
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I know, I know, people don't like the color scheme I use, but I got so
used to it and with the new version generated via ColorTemplates it
messes with me being used to elflord with 16 terminal colors instead of
the GUI colors.
So this patches elflord in such a way that it's consistent with what I
want but still uses the new ColorTemplates version, which for example
comes with nicer colors for vimdiff.
Signed-off-by: aszlig <aszlig@nix.build>
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
Nope, never mind, maple is still not building, therefore vim-clap can't
build.
This reverts commit 4d4f07ef82d8a19587cc450b6d0a198e16acb48f.
|
|
|
|
| |
Apparently this builds again according to `hydra-check`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The patch file was relative to the rustfmt crate within the rustc source
tree, so I added "-d src/tools/rustfmt" as patchFlags to apply the patch
against that subdirectory.
Unfortunately, patchFlags also applies to other patches as well and I
didn't account for that. With current rustfmt from rustc 1.59.0 we have
a patch in nixpkgs, which now fails to apply because of this:
applying patch /nix/store/...-rustfmt-fix-self-tests.patch
can't find file to patch at input line 3
Perhaps you used the wrong -p or --strip option?
The text leading up to this was:
--------------------------
|--- a/src/tools/rustfmt/src/ignore_path.rs
|+++ b/src/tools/rustfmt/src/ignore_path.rs
--------------------------
File to patch:
Skip this patch? [y]
Skipping patch.
1 out of 1 hunk ignored
To fix this, I changed the paths in config.patch to be relative to the
rustc tree rather than just the rustfmt crate, so the patch should now
apply regardless of whether there are other patches.
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
| |
*sigh*
|
|
|
|
|
|
| |
* plumb/plumber for mouse 3
* rc for mouse 2
* win for terminal feature
|
| |
|
|
|
|
|
|
|
| |
The patch no longer applies because some of the imports got changed in
the context of the diff.
Signed-off-by: aszlig <aszlig@nix.build>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The version from 2021 certainly is way too old for my desire to crave
new bugs, so let's update it for no particular reason other than gut
feeling.
Oh, well and of course I saw a bunch of OMEMO plugin fixes, so maybe if
you search hard enough you'll find compelling reasons... :-D
Signed-off-by: aszlig <aszlig@nix.build>
|
|
|
|
|
|
|
| |
nix_2_3 is an alias for nixVersions.nix_2_3 upstream now and the test
infrastructure from nixpkgs seems to default to allowAliases = false;
when evaluating (which is actually quite concerning for downstream
usage).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
depot is a nix-based monorepo which contains some great nix utilities
like yants (a nix type system), runTestsuite, mergePatch and so on, a
few interesting pure nix builders like buildLisp and buildGo and a few
packages maintained by @Profpatsch and myself.
This change exposes tvl completely as pkgs.tvl, but prevents hydra from
building it using dontRecurseIntoAttrs as depot pins its own version of
nixpkgs which is not easily overrideable, contains some expensive to
build system configurations we are not interested in and even some
notoriously indeterministic packages.
Additionally it is possible to override pkgs.tvl to use a different or
local version of depot:
pkgs.tvl.override { tvlSrc = /home/lukas/src/depot; }
To keep with @Profpatsch's previous solution, we pass in vuizvui's
nixpkgs version to depot via nixpkgsBisectPath which may break packages
in depot occasionally if nixpkgs causes breakage in TVL and depot isn't
updated accordingly.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Patch plan9port most notably including a patch that fixes the
interpretation of X11 key events. This resolves the problem that using
the shift key of the neo layout would cause acme to start interpreting
left mouse button as right mouse button.
* Add a wrapper derivation that puts acme and all executables it
absolutely requires in PATH. The full Plan 9 userland can still be
accessed via the 9 executable, but this allows executing just acme and
access e.g. the GNU coreutils from inside of it — in case you do want
to use cat -v.
machines/sternenseemann/wolfgang: install acme
|