enhance(main/neovim{,-nightly},main/vim,x11/vim-gtk): document `has('termux')` feature flag by TomJo2000 · Pull Request #28681 · termux/termux-packages

TomJo2000 · 2026-03-04T00:43:41Z

This PR replaces the src-nvim-eval-funcs.c.patch for the neovim/neovim-nightly packages with a more comprehensive version, has-termux.patch.

The renamed and expanded patch adds documentation of the has('termux') and has('android') feature flags added by the original patch.
In addition the added documentation also makes available the :h termux and :h android help tags, pointing to the additional information about the feature flags.

The syntax and style guide for the vimdoc documentation format are documented in:

:h help-writing and
:h dev-doc

The formatting on the help file is a bit wonky in the patch, however I have confirmed manually that it is indented identically to the rest of the section.

^{(Contrast of indentation guides enhanced manually for visual clarity.)}

TomJo2000 · 2026-03-04T01:14:47Z

The build currently seems to die on this right here.

termux-packages/build-package.sh

Lines 652 to 662 in f9bc63a

    
           for (( i=0; i < ${#PACKAGE_LIST[@]}; i++ )); do 
        
           	# Following commands must be executed under lock to prevent running 
        
           	# multiple instances of "./build-package.sh". 
        
           	# 
        
           	# To provide a sane environment for each package, 
        
           	# builds are done in an explicit subshell for each. 
        
           	# shellcheck disable=SC2031 
        
           	( 
        
           		if [[ "$TERMUX_BUILD_IGNORE_LOCK" != "true" ]]; then 
        
           			flock -n 5 || termux_error_exit "Another build is already running within same environment." 
        
           		fi

Which would indicate that the flock is not being properly released in this instance.
I can replicate the issue locally so this is probably not directly related to AppArmor.

TomJo2000 · 2026-03-04T01:35:29Z

I have a feeling we accidentally disallowed the container from using the flock() syscall, or a related one that's required here.

That would have been introduced by 221ce1f#diff-6f419dc9863f442f8b508a38b040ec549d82031311653f4b372243adc3653987, but I don't know how I could test that to verify.

More eyes would be appreciated on this.
CC: @thunder-coding @robertkirkman

robertkirkman · 2026-03-04T02:56:40Z

I don't have apparmor so I'm not sure if I will be a lot of help in this case, but thunder-coding does and will be able to locally troubleshoot this.

robertkirkman · 2026-03-04T02:57:26Z

I can replicate the issue locally so this is probably not directly related to AppArmor.

You are running apparmor locally as well, though, aren't you?

TomJo2000 · 2026-03-04T03:02:17Z

You are running apparmor locally as well, though, aren't you?

Nope.
No AppArmor locally, hence why I didn't think it was directly AppArmor related.

robertkirkman · 2026-03-04T03:37:00Z

Oh, I misunderstood and I thought you had AppArmor. I'll try to reproduce this

robertkirkman · 2026-03-04T03:44:46Z

I wasn't able to reproduce this when just using scripts/run-docker.sh ./build-package.sh -I -f neovim neovim-nightly in an up-to-date repository folder

TomJo2000 · 2026-03-04T04:15:21Z

I wasn't able to reproduce this when just using scripts/run-docker.sh ./build-package.sh -I -f neovim neovim-nightly in an up-to-date repository folder

Is your docker container up to date?

thunder-coding · 2026-03-04T04:18:01Z

Which would indicate that the flock is not being properly released in this instance.
I can replicate the issue locally so this is probably not directly related to AppArmor.

I was able to replicate this locally in some instances, but during later testing they went away, so I thought it was just some weird codepath that didn't trigger when I completed my PR. And this is affecting CI as well now. So will try digging into after python rebuilds are complete

robertkirkman · 2026-03-04T04:20:59Z

I wasn't able to reproduce this when just using scripts/run-docker.sh ./build-package.sh -I -f neovim neovim-nightly in an up-to-date repository folder

Is your docker container up to date?

Yes

licy183 · 2026-03-04T15:55:01Z

It may not be related to flock, maybe just because that fuse-overlayfs has not stopped running.

thunder-coding · 2026-03-04T20:18:49Z

4fd4e48 should fix this

TomJo2000 · 2026-03-04T23:05:10Z

Just checked the DEBs.

Seeing a distinct lack of Termux in that documentation.
Will need to look into why it's not getting generated properly.

TomJo2000 · 2026-03-05T04:39:20Z

We need to explicitly run the make doc target.
The docs are usually pre-generated as part of the release process.

TomJo2000 · 2026-03-05T05:24:32Z

/home/builder/.termux-build/neovim/src/.deps/build/src/libuv/src/unix/linux.c:2331:36: error: use of undeclared identifier 'LLONG_MAX'
 2331 |     constraint->quota_per_period = LLONG_MAX;
      |                                    ^
1 error generated.
gmake[6]: *** [CMakeFiles/uv_a.dir/build.make:513: CMakeFiles/uv_a.dir/src/unix/linux.c.o] Error 1
gmake[6]: Leaving directory '/home/builder/.termux-build/neovim/src/.deps/build/src/libuv-build'
gmake[5]: *** [CMakeFiles/Makefile2:87: CMakeFiles/uv_a.dir/all] Error 2
gmake[5]: Leaving directory '/home/builder/.termux-build/neovim/src/.deps/build/src/libuv-build'
gmake[4]: *** [Makefile:136: all] Error 2
gmake[4]: Leaving directory '/home/builder/.termux-build/neovim/src/.deps/build/src/libuv-build'
gmake[3]: *** [CMakeFiles/libuv.dir/build.make:86: build/src/libuv-stamp/libuv-build] Error 2
gmake[3]: Leaving directory '/home/builder/.termux-build/neovim/src/.deps'
gmake[2]: *** [CMakeFiles/Makefile2:158: CMakeFiles/libuv.dir/all] Error 2
gmake[2]: Leaving directory '/home/builder/.termux-build/neovim/src/.deps'
gmake[1]: *** [Makefile:91: all] Error 2
gmake[1]: Leaving directory '/home/builder/.termux-build/neovim/src/.deps'
make: *** [Makefile:108: deps] Error 2

Yeah I got nothing...
No idea how running make doc causes that to suddenly fail.

TomJo2000 · 2026-03-05T05:39:45Z

Moving it to termux_step_post_configure() makes it fail on Luajit instead.

Apparently running make doc causes all sorts of unexplained havoc.

TomJo2000 · 2026-03-06T17:38:32Z

Moving the make doc to termux_step_host_build() was also unsuccessful.

We can just patch the pre-generated docs in the tarball, it's fine.
It's not what I'd prefer, but it should be equally resilient against patch drift.

justinmk · 2026-03-09T18:15:56Z

+---   android    Alias for has('termux')
+---   termux    Termux environment (The *termux* and *android*


Not an objection, just curious: does vim.uv.os_uname() return enough info to achieve this?

I don't think that's gonna be terribly useful unfortunately.

-- package version: neovim 0.11.6-2 -- nvim --clean { machine = "aarch64", release = "5.4.274-qgki-g6ef38bd89f5c", sysname = "Linux", version = "#1 SMP PREEMPT Wed Jan 28 16:01:13 CST 2026" }

I get the same result with neovim-nightly (0.12.0~dev-2508+g61f166ec40)

Anything else come to mind that could be a useful proxy?
For compile time Android checking we usually use #ifdef __ANDROID__ and/or __TERMUX__.

We should be able to apply that to the src/nvim/eval/funcs.c part of this patch at least.
Which is taken from this prior original patch
^{https://github.com/termux/termux-packages/blob/994dcc0/packages/neovim/src-nvim-eval-funcs.c.patch}

Originally introduced in Make has(android/termux) return true in neovim/vim #7545

An equivalent patch also exists for our vim/vim-gtk package, so I should at least update that patch to include documentation of the feature flags as well.

I'm not sure how we could cleanly extrapolate the presence of those definitions from the C preprocessor to the doc generation,
or if we can use an equivalent proxy for that.

We could add a build flag so that has('foo') works for an arbitrary build-time value, but yeah the docs are less obvious.

I'm working on updating the vim patch right now,
but yeah I'm still drawing a blank for making that conditional for the docs.

actually, we can just update the docs directly upstream. PR would be appreciated

mention this issue/PR so we have a reference

Guess I'm making a PR to Vim and Neovim then.
I just finished reworking the vim patch as well.

I'm guessing specific wording can be worked out there, what I have here is just the first things that came to mind.

With the C pre-proc macro checks this also decouples the termux and android feature flags.
So if some other project wants to compile Vim/Neovim on Android they wouldn't necessarily inherit the termux feature, since the __TERMUX__ macro is added as part of out Android NDK patches.

termux-packages/ndk-patches/29/sys-cdefs.h.patch

Lines 1 to 9 in 0c87916

--- ./usr/include/sys/cdefs.h.orig 2021-08-16 11:52:08.128876161 +0200

+++ ./usr/include/sys/cdefs.h 2021-08-16 11:52:08.555542676 +0200

@@ -372,3 +372,6 @@

#if __has_include(<android/ndk-version.h>)

#include <android/ndk-version.h>

#endif

+#define __TERMUX__ 1

+#define __TERMUX_PREFIX__ "@TERMUX_PREFIX@"

+

…ture flag

TomJo2000 · 2026-03-18T08:26:27Z

Superseded by:

TomJo2000 force-pushed the nvim-has-termux-docs branch 2 times, most recently from 52b92df to 50e987f Compare March 4, 2026 01:11

thunder-coding self-assigned this Mar 4, 2026

thunder-coding added this to @thunder-coding's public tracker Mar 4, 2026

thunder-coding moved this to Todo in @thunder-coding's public tracker Mar 4, 2026

thunder-coding approved these changes Mar 4, 2026

View reviewed changes

TomJo2000 force-pushed the nvim-has-termux-docs branch from 50e987f to e91addb Compare March 4, 2026 20:21

TomJo2000 force-pushed the nvim-has-termux-docs branch from e91addb to fd88be6 Compare March 5, 2026 04:47

TomJo2000 force-pushed the nvim-has-termux-docs branch 2 times, most recently from dc5c67a to 760c3e8 Compare March 6, 2026 17:36

TomJo2000 force-pushed the nvim-has-termux-docs branch 3 times, most recently from 27f4b8e to 978717b Compare March 9, 2026 16:26

TomJo2000 mentioned this pull request Mar 9, 2026

fix(cmake): use host nlua0 binary when cross-compiling neovim/neovim#38089

Merged

justinmk reviewed Mar 9, 2026

View reviewed changes

TomJo2000 added 2 commits March 9, 2026 23:38

enhance(main/neovim{,-nightly}): document has('termux') feature flag

8975eb7

squash! enhance(main/neovim{,-nightly}): document has('termux') fea…

b356259

…ture flag

TomJo2000 force-pushed the nvim-has-termux-docs branch from 978717b to b356259 Compare March 9, 2026 22:39

TomJo2000 changed the title ~~enhance(main/neovim{,-nightly}): document has('termux') feature flag~~ enhance(main/neovim{,-nightly},main/vim,x11/vim-gtk): document has('termux') feature flag Mar 9, 2026

enhance(main/vim,x11/vim-gtk): document has('termux') feature flag

ac2ca0c

This was referenced Mar 10, 2026

feat(has): 'android' and 'termux' feature flag neovim/neovim#38218

Merged

feat(has): add support for 'android' and 'termux' feature flags vim/vim#19623

Closed

TomJo2000 marked this pull request as draft March 14, 2026 04:30

TomJo2000 closed this Mar 18, 2026

github-project-automation Bot moved this from Todo to Done in @thunder-coding's public tracker Mar 18, 2026

TomJo2000 deleted the nvim-has-termux-docs branch March 18, 2026 08:27

		+--- android Alias for has('termux')
		+--- termux Termux environment (The termux and android

	--- ./usr/include/sys/cdefs.h.orig 2021-08-16 11:52:08.128876161 +0200
	+++ ./usr/include/sys/cdefs.h 2021-08-16 11:52:08.555542676 +0200
	@@ -372,3 +372,6 @@
	#if __has_include(<android/ndk-version.h>)
	#include <android/ndk-version.h>
	#endif
	+#define __TERMUX__ 1
	+#define __TERMUX_PREFIX__ "@TERMUX_PREFIX@"
	+

Uh oh!

Conversation

TomJo2000 commented Mar 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

The syntax and style guide for the vimdoc documentation format are documented in:

Uh oh!

TomJo2000 commented Mar 4, 2026

Uh oh!

TomJo2000 commented Mar 4, 2026

Uh oh!

robertkirkman commented Mar 4, 2026

Uh oh!

robertkirkman commented Mar 4, 2026

Uh oh!

TomJo2000 commented Mar 4, 2026

Uh oh!

robertkirkman commented Mar 4, 2026

Uh oh!

robertkirkman commented Mar 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

TomJo2000 commented Mar 4, 2026

Uh oh!

thunder-coding commented Mar 4, 2026

Uh oh!

robertkirkman commented Mar 4, 2026

Uh oh!

licy183 commented Mar 4, 2026

Uh oh!

thunder-coding commented Mar 4, 2026

Uh oh!

TomJo2000 commented Mar 4, 2026

Uh oh!

TomJo2000 commented Mar 5, 2026

Uh oh!

TomJo2000 commented Mar 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

TomJo2000 commented Mar 5, 2026

Uh oh!

TomJo2000 commented Mar 6, 2026

Uh oh!

justinmk Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

TomJo2000 Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

TomJo2000 Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

justinmk Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

TomJo2000 Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

justinmk Mar 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

TomJo2000 Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

TomJo2000 commented Mar 18, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

TomJo2000 commented Mar 4, 2026 •

edited

Loading

robertkirkman commented Mar 4, 2026 •

edited

Loading

TomJo2000 commented Mar 5, 2026 •

edited

Loading

justinmk Mar 9, 2026 •

edited

Loading