feat: add FreeBSD support #7606

charlesrocket · 2025-06-16T16:39:35Z

charlesrocket · 2025-06-16T17:06:31Z

libxev needs to be replaced after the upstream merge

pluiedev

Also in general I think that instead of Linux/BSD we should just unify on "the GTK app", even though GTK can technically also be built for macOS and Windows. It's a bit of a mess, since FreeBSD behaves like Linux in userland and more like macOS in kernel land, so I really am not sure how best we should explain this.

build.zig

pluiedev · 2025-06-16T20:26:23Z

src/config/Config.zig

@@ -1823,7 +1823,7 @@ keybind: Keybinds = .{},
 /// Automatically hide the quick terminal when focus shifts to another window.
 /// Set it to false for the quick terminal to remain open even when it loses focus.
 ///
-/// Defaults to true on macOS and on false on Linux. This is because global
+/// Defaults to true on macOS and on false on Linux/BSD. This is because global
 /// shortcuts on Linux require system configuration and are considerably less


You forgot to edit this one, too. I wonder if we should just replace all instances of "Linux" to "the GTK app (runtime)"... pending discussion

Another option is to replace Linux with Unix/Unix-like.

The thing is that macOS is also technically a Unix system (even more so than Linux), but on the user-facing level the BSDs are far more similar to Linux. It's a tricky nomenclature problem for sure

pluiedev · 2025-06-16T20:27:10Z

src/config/Config.zig

+@"linux-cgroup": LinuxCgroup = if (builtin.os.tag == .linux)
+    .@"single-instance"
+else
+    .never,


Why do we need a specific switch for Linux even though the option already indicates that it's Linux-specific? Maybe we should rename the option then?

It should be just cgroups since one can enable it on BSD as well, so I just gated it for now. Config change such as this should probably be done in its own PR, but I can push it if it is okay.

src/os/i18n.zig

pluiedev · 2025-06-16T20:59:35Z

I'm going to convert this PR to draft for now since some parts of it are still unfinished (i.e. localization support) and pending on the libxev PR. It's just not productive right now IMO to assess this PR as it currently stands. Once the libxev PR is merged and this PR has reached (near-)feature parity with Linux, then we can again mark it as ready for review.

charlesrocket · 2025-06-16T21:20:39Z

@pluiedev fixed local issues, builds are passing now with i18n. Oh, theres error on the release that I need to look into.

tristan957 · 2025-06-16T21:33:46Z

.github/workflows/test.yml

+          run: |
+            # switch to pkg
+            wget https://ziglang.org/builds/zig-x86_64-freebsd-0.15.0-dev.777+6810ffa42.tar.xz
+            tar -xf zig-x86_64-freebsd-0.15.0-dev.777+6810ffa42.tar.xz
+            zig-x86_64-freebsd-0.15.0-dev.777+6810ffa42/zig build -Dapp-runtime=none test


Hmm, is it not possible to use 0.14.0 or 0.14.1? Seems like this could cause headaches later.

Fresh zig is already in the latest branch so it will be in quarterly soon. This is just a temp workaround.

00-kat · 2025-06-16T23:41:18Z

Also in general I think that instead of Linux/BSD we should just unify on "the GTK app"

Aside: the repo currently has os/macos and os/linux labels; would FreeBSD get its own label or should there be a label for “GTK app” instead?

charlesrocket · 2025-06-17T00:40:15Z

@00-kat I'd say it should be both BSD and GTK (gui/gtk). There is probably no need to have individual labels for BSDs (at least not yet).

charlesrocket · 2025-06-17T12:19:15Z

Translations are working as expected now.

mitchellh

Looks very good. Here is the TODO on my end:

Review the xev changes and fix as necessary
Migrate CI to Namespace and verify it works there
Consider adding prefer to the non-dynamic xev to avoid the comptime checks here

src/build/GhosttyI18n.zig

mitchellh · 2025-06-17T20:19:27Z

Note: vmactions/freebsd-vm doesn't work on Namespace yet. I've opened a support ticket with them to get this looked at (cc @hugosantos). I didn't block libxev on it because its low traffic, but I want to wait for that to merge this because its so much more expensive to run Ghostty's CI.

charlesrocket · 2025-06-17T22:14:00Z

@mitchellh I thought boxybsd could be used, but I doubt I can get GH runner working. There is also https://github.com/acj/freebsd-firecracker-action.

mitchellh · 2025-06-18T15:28:20Z

Namespace fixed the issue, we can run the FreeBSD VM action on Namespace now, see libxev: mitchellh/libxev#168

mitchellh · 2025-06-18T15:30:28Z

Moving this to ready for review and will make some tweaks to get this merged. It may not be perfect yet but it at least gets Ghostty working to some extent (similar to Windows).

mitchellh · 2025-06-18T15:35:00Z

@charlesrocket We only test unit tests on FreeBSD right now in CI. It would be better to test both the GLFW and GTK builds as well. Can you work on a future PR to address that?

mitchellh

Excellent work @charlesrocket. Thank you very much! 😄

FreeBSD will remain a lower tier platform for now since none of the maintainers have access to actually test on it, but we at least have CI ensuring we don't break it and an active community member to help support it.

We can talk about more maintenance going forward in Discord. Thanks 😄

mitchellh · 2025-06-18T15:43:29Z

Hm, looks like the FreeBSD CI job failed... kind of a weird failure. I'll rerun it and see if that works.

charlesrocket · 2025-06-18T15:45:30Z

@mitchellh definitely, finishing the freebsd port and will switch to GLFW build error.

mitchellh · 2025-06-21T14:21:50Z

@charlesrocket The firecracker VM approach didn't work for probably unrelated CI reasons. I'm sorry I haven't had time to look into this yet. I'd be fine merging this WITHOUT CI for now but we have to both be okay with the reality that FreeBSD support will probably bitrot until we have CI support because we don't have anyone staying on top of it.

If you're okay with that in the interim we can remove the freebsd job and try a subsequent PR to add it.

charlesrocket · 2025-06-21T16:23:36Z

@mitchellh I am okay with missing CI for now

pluiedev

I still have my reservations about some of these changes (a lot of them look a bit too tacked-on with the builtin.target.os.tag check), but I don't feel like blocking this further.

In the future I think there needs to be a cleanup PR to unify the terminology a little bit: we should use "GTK app" instead of "Linux" in a lot of cases within the codebase, unless something is truly specific to Linux (nothing immediately comes to mind yet). This IMO is the least intrusive option since although the GTK app can technically be built for macOS, we never distribute it in that form and we generally assume it to be run in a FreeDesktop-based environment (Linux, FreeBSD, etc.). We also should probably update the website ahead of Ghostty 1.2 to reflect this nomenclature change.

pluiedev · 2025-06-21T18:07:40Z

src/build/GhosttyI18n.zig

@@ -21,14 +22,22 @@ pub fn init(b: *std.Build, cfg: *const Config) !GhosttyI18n {
    defer steps.deinit();

    inline for (internal_os.i18n.locales) |locale| {
+        // There is no encoding suffix in the LC_MESSAGES path on FreeBSD,


Not a blocker, but I'm curious how programs like vim gets away with it on FBSD. Are there truly no programs out there that contain the encoding suffix?

pluiedev · 2025-06-21T18:09:19Z

src/config/Config.zig

@@ -1823,7 +1823,7 @@ keybind: Keybinds = .{},
 /// Automatically hide the quick terminal when focus shifts to another window.
 /// Set it to false for the quick terminal to remain open even when it loses focus.
 ///
-/// Defaults to true on macOS and on false on Linux. This is because global
+/// Defaults to true on macOS and on false on Linux/BSD. This is because global
 /// shortcuts on Linux require system configuration and are considerably less


Note to self: go through all the configs mentioning "Linux" and think about whether it applies to BSDs as well

pluiedev · 2025-06-21T18:15:19Z

src/config/Config.zig

@@ -2812,7 +2815,7 @@ pub fn loadCliArgs(self: *Config, alloc_gpa: Allocator) !void {
    //     styling, etc. based on the command.
    //
    // See: https://github.com/Vladimir-csp/xdg-terminal-exec
-    if (comptime builtin.os.tag == .linux) {
+    if ((comptime builtin.os.tag == .linux) or (comptime builtin.os.tag == .freebsd)) {


We should probably add a helper function inside std/os that determines whether the current system is based on FreeDesktop (XDG) standards, since I can totally see us needing to expand this check a lot later on.

Owl-Lord · 2025-06-21T20:24:23Z

i mean if the want for more portability beyond freebsd is interesting i have quiet a few friends who use openbsd which i could ask to help with knowledge about the inner workings of it

mitchellh · 2025-06-21T21:06:38Z

@pluiedev

I still have my reservations about some of these changes (a lot of them look a bit too tacked-on with the builtin.target.os.tag check), but I don't feel like blocking this further.

I don't think there was much. There was zero cases added to the GTK app (just an xev check unrelated to BSD), and almost the entirety of OS tag checks are in the build system and src/os package, both of which are inherently extremely OS/target dependent. I think it's pretty clean, considering.

In the future I think there needs to be a cleanup PR to unify the terminology a little bit: we should use "GTK app" instead of "Linux" in a lot of cases within the codebase, unless something is truly specific to Linux (nothing immediately comes to mind yet).

This is probably right. I've avoided "GTK" in case we ever actually dropped GTK as the default for Linux (and now BSD). But that's probably a stretch. Using the apprt as the description is probably more appropriate. I think macOS is in line with that because while macOS uses the libghostty apprt technically, its really a dedicated macOS app there consuming that.

Following cryptocode@7aeadb0

Temp

Waiting for mitchellh/libxev#167

for now

mitchellh

Approved, given the caveats noted earlier, which I'll repeat here.

We never got CI working. As such, we've all agreed that the FreeBSD support may bitrot. We don't have any active maintainers running FreeBSD (or any BSD for that matter). We can't guarantee that PRs the community or ourselves make will continue to work on FreeBSD. As such, we should strive to get FreeBSD in CI.

Until then, I don't want to block improvements on our portability, so I'm willing to approve and merge with that disclaimer out of the way.

And I want to thank @charlesrocket again for the excellent effort and work to get us here!

charlesrocket requested a review from a team as a code owner June 16, 2025 16:39

pluiedev requested changes Jun 16, 2025

View reviewed changes

pluiedev marked this pull request as draft June 16, 2025 20:59

charlesrocket force-pushed the fbsd branch from f0e3659 to 8b8b4cb Compare June 16, 2025 21:19

tristan957 reviewed Jun 16, 2025

View reviewed changes

charlesrocket force-pushed the fbsd branch from 8b8b4cb to 6a12b18 Compare June 17, 2025 12:16

mitchellh reviewed Jun 17, 2025

View reviewed changes

src/build/GhosttyI18n.zig Show resolved Hide resolved

mitchellh marked this pull request as ready for review June 18, 2025 15:30

mitchellh force-pushed the fbsd branch from ba5ff36 to 25ed1d4 Compare June 18, 2025 15:34

mitchellh approved these changes Jun 18, 2025

View reviewed changes

charlesrocket requested a review from a team as a code owner June 19, 2025 11:28

charlesrocket force-pushed the fbsd branch 2 times, most recently from 658e6a9 to 4a5164b Compare June 19, 2025 21:54

pluiedev approved these changes Jun 21, 2025

View reviewed changes

charlesrocket and others added 23 commits June 21, 2025 14:11

Add FreeBSD support

e09657e

Following cryptocode@7aeadb0

Import libxev fork

c1830b5

Temp

test: fix desktop cases on BSD

e2f86f0

fix: get GTK env var on FreeBSD

e293744

fix: check DE env var on FreeBSD

8e7e9cb

fix(config): fix quick-terminal-autohide comment

6e190ac

fix(config): enable cgroups for linux only

baa41c3

ci: add freebsd job

4510433

build(deps): bump charlesrocket/libxev to e29fc0c

4d7a667

Waiting for mitchellh/libxev#167

test: fix desktop cases

3e582a6

build: disable i18n on FreeBSD

780ee89

for now

fix: enable i18n on FreeBSD

1634854

style: fix formatting

43b8472

build: comment locale trim

a209494

fix: update libxev

d6b2d0e

ci: run freebsd tests on Namespace

2314d3d

update nix hash

c4d5949

flatpak: update hash

43a46d1

build: fix fontconfig paths on FreeBSD

89b3977

build: fix terminfo location on FreeBSD

8b0de9b

fix: set resources_dir on FreeBSD

b32b1e7

ci: switch to freebsd-firecracker-action

097d1ad

ci: remove freebsd test for now

888daca

mitchellh force-pushed the fbsd branch from ccce278 to 888daca Compare June 21, 2025 21:12

mitchellh approved these changes Jun 21, 2025

View reviewed changes

mitchellh merged commit 9d922e1 into ghostty-org:main Jun 21, 2025
37 checks passed

github-actions bot added this to the 1.2.0 milestone Jun 21, 2025

feat: add FreeBSD support #7606

feat: add FreeBSD support #7606

Uh oh!

Conversation

charlesrocket commented Jun 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

charlesrocket commented Jun 16, 2025

Uh oh!

pluiedev left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

charlesrocket Jun 16, 2025 • 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!

Uh oh!

pluiedev commented Jun 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

charlesrocket commented Jun 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

00-kat commented Jun 16, 2025

Uh oh!

charlesrocket commented Jun 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

charlesrocket commented Jun 17, 2025

Uh oh!

mitchellh left a comment • 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!

Uh oh!

mitchellh commented Jun 17, 2025

Uh oh!

charlesrocket commented Jun 17, 2025

Uh oh!

mitchellh commented Jun 18, 2025

Uh oh!

mitchellh commented Jun 18, 2025

Uh oh!

mitchellh commented Jun 18, 2025

Uh oh!

mitchellh left a comment

Choose a reason for hiding this comment

Uh oh!

mitchellh commented Jun 18, 2025

Uh oh!

charlesrocket commented Jun 18, 2025

Uh oh!

mitchellh commented Jun 21, 2025

Uh oh!

charlesrocket commented Jun 21, 2025

Uh oh!

pluiedev left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Owl-Lord commented Jun 21, 2025

Uh oh!

charlesrocket commented Jun 16, 2025 •

edited

Loading

charlesrocket Jun 16, 2025 •

edited

Loading

pluiedev commented Jun 16, 2025 •

edited

Loading

charlesrocket commented Jun 16, 2025 •

edited

Loading

charlesrocket commented Jun 17, 2025 •

edited

Loading

mitchellh left a comment •

edited

Loading

mitchellh left a comment •

edited

Loading