| 19 Dec 2025 |
 dotlambda | https://github.com/NixOS/nixpkgs/pull/472457 | 19:44:39 |
 ghpzin | Noticed that all graphical xserver tests seem to be broken on staging-next.
wayland is fine, tests for gnome,sway,xfce-wayland, cinnamon-wayland (and plasma6 if you switch test to wayland) pass.
lightdm does not tell much, just that display-manager.service fails:
systemd[1]: display-manager.service: Main process exited, code=exited, status=1/FAILURE
systemd[1]: display-manager.service: Failed with result 'exit-code'.
sddm and ly in tests show errors from xorg-server:
xserver-wrapper[823]: (II) LoadModule: "modesetting"
xserver-wrapper[823]: (II) Loading /nix/store/lrbawv4ccl0x12nwrmbbzpxmzjj1z5r6-xorg-server-21.1.20/lib/xorg/modules/drivers/modesetting_drv.so
xserver-wrapper[823]: (EE) Failed to load /nix/store/lrbawv4ccl0x12nwrmbbzpxmzjj1z5r6-xorg-server-21.1.20/lib/xorg/modules/drivers/modesetting_drv.so: /nix/store/lrbawv4ccl0x12nwrmbbzpxmzjj1z5r6-xorg-server-21.1.20/lib/xorg/modules/drivers/modesetting_drv.so: undefined symbol: gbm_bo_get_plane_count
xserver-wrapper[823]: (EE) Failed to load module "modesetting" (loader failed, 0)
xserver-wrapper[823]: (EE) No drivers available.
Changing this to "--disable-glamor" fixes it:
https://github.com/NixOS/nixpkgs/blob/fd720f4fd4e78cce030a72026d1b7cadf6454f29/pkgs/by-name/xo/xorg-server/package.nix#L185
I thought it may be related to 21.1.18 -> 21.1.20 update, similar to:
https://github.com/NixOS/nixpkgs/issues/471620
but reverting version to 21.1.18 does not help, neither does updating to 21.1.21
There were a few xorg refactors merged to staging, maybe one of them did something unintentional. | 19:52:10 |
 K900 | Can you check if modesetting_drv links libgbm correctly | 20:15:25 |
| K900 | gbmbogetplanecount should be exported by libgbm | 20:15:47 |
| ghpzin | It doesn't link directly right, something else there should link to that ?
Unless I misunderstand what you mean, ie on both staging-next and master it does not link to libgbm:
ldd ./result-master/lib/xorg/modules/drivers/modesetting_drv.so
linux-vdso.so.1 (0x00007fb435c72000)
libudev.so.1 => /nix/store/q5fmjr9s9968bkv5bm3al5517x27bm19-systemd-minimal-libs-258.2/lib/libudev.so.1 (0x00007fb435b03000)
libm.so.6 => /nix/store/xx7cm72qy2c0643cm1ipngd87aqwkcdp-glibc-2.40-66/lib/libm.so.6 (0x00007fb435a1b000)
libc.so.6 => /nix/store/xx7cm72qy2c0643cm1ipngd87aqwkcdp-glibc-2.40-66/lib/libc.so.6 (0x00007fb435800000)
libcap.so.2 => /nix/store/l486byj55nq10ys3jbgszd2hywkxbq7l-libcap-2.77-lib/lib/libcap.so.2 (0x00007fb435a0e000)
/nix/store/j193mfi0f921y0kfs8vjc1znnr45ispv-glibc-2.40-66/lib64/ld-linux-x86-64.so.2 (0x00007fb435c74000)
The only thing in both cases that does is /lib/xorg/modules/libglamoregl.so:
ldd ./result-master/lib/xorg/modules/libglamoregl.so
linux-vdso.so.1 (0x00007fa0653d5000)
libgbm.so.1 => /nix/store/pw8v6hlrg7lz2d3x115x3zy80bbhx1x9-mesa-libgbm-25.1.0/lib/libgbm.so.1 (0x00007fa065344000)
libepoxy.so.0 => /nix/store/1r61lw1k35085way0cgl3pxz9cm9awjs-libepoxy-1.5.10/lib/libepoxy.so.0 (0x00007fa06521e000)
libm.so.6 => /nix/store/xx7cm72qy2c0643cm1ipngd87aqwkcdp-glibc-2.40-66/lib/libm.so.6 (0x00007fa065136000)
libc.so.6 => /nix/store/xx7cm72qy2c0643cm1ipngd87aqwkcdp-glibc-2.40-66/lib/libc.so.6 (0x00007fa064e00000)
libdrm.so.2 => /nix/store/p4fk8qb6cy0zs65j6h50rh9bs7hnsgif-libdrm-2.4.128/lib/libdrm.so.2 (0x00007fa06511b000)
/nix/store/j193mfi0f921y0kfs8vjc1znnr45ispv-glibc-2.40-66/lib64/ld-linux-x86-64.so.2 (0x00007fa0653d7000)
ldd ./result-staging-next/lib/xorg/modules/libglamoregl.so
linux-vdso.so.1 (0x00007f8e1c2d2000)
libgbm.so.1 => /nix/store/24w3s75aa2lrvvxsybficn8y3zxd27kp-mesa-libgbm-25.1.0/lib/libgbm.so.1 (0x00007f8e1c23f000)
libepoxy.so.0 => /nix/store/kzl2j4nxwr6c0f2xbd44zjmgh8kqxr57-libepoxy-1.5.10/lib/libepoxy.so.0 (0x00007f8e1c11c000)
libm.so.6 => /nix/store/j193mfi0f921y0kfs8vjc1znnr45ispv-glibc-2.40-66/lib/libm.so.6 (0x00007f8e1c034000)
libc.so.6 => /nix/store/j193mfi0f921y0kfs8vjc1znnr45ispv-glibc-2.40-66/lib/libc.so.6 (0x00007f8e1be00000)
libdrm.so.2 => /nix/store/sixic8a4adxz72swww4fcls0h5sabjdf-libdrm-2.4.129/lib/libdrm.so.2 (0x00007f8e1c019000)
/nix/store/j193mfi0f921y0kfs8vjc1znnr45ispv-glibc-2.40-66/lib64/ld-linux-x86-64.so.2 (0x00007f8e1c2d4000)
| 20:34:48 |
| K900 | Hm | 20:36:34 |
| K900 | I wonder if it's the xserver itself then | 20:36:40 |
| K900 | That is linking it | 20:36:46 |
| K900 | Is it just a load order thing | 20:37:04 |
| K900 | That would be sad | 20:37:07 |
| ghpzin | I tried reordering these buildInputs like here on master:
https://github.com/NixOS/nixpkgs/blob/c30b80416b233c3673e573c0821bf96cc83818db/pkgs/servers/x11/xorg/overrides.nix#L452-L457
ie this diff on staging-next:
--- a/pkgs/by-name/xo/xorg-server/package.nix
+++ b/pkgs/by-name/xo/xorg-server/package.nix
@@ -139,10 +139,10 @@
xtrans
]
++ lib.optionals (!stdenv.hostPlatform.isDarwin) [
- dri-pkgconfig-stub
libdrm
libgbm
mesa-gl-headers
+ dri-pkgconfig-stub
]
++ lib.optionals stdenv.hostPlatform.isDarwin [
autoconf
but it did not help.
That refactor moved xorg-server into separate by-name so tracking what went where is quite hard. | 20:42:35 |
| ghpzin | I tried reordering these buildInputs like here on master:
https://github.com/NixOS/nixpkgs/blob/c30b80416b233c3673e573c0821bf96cc83818db/pkgs/servers/x11/xorg/overrides.nix#L452-L457
ie this diff on staging-next:
--- a/pkgs/by-name/xo/xorg-server/package.nix
+++ b/pkgs/by-name/xo/xorg-server/package.nix
@@ -139,10 +139,10 @@
xtrans
]
++ lib.optionals (!stdenv.hostPlatform.isDarwin) [
- dri-pkgconfig-stub
libdrm
libgbm
mesa-gl-headers
+ dri-pkgconfig-stub
]
++ lib.optionals stdenv.hostPlatform.isDarwin [
autoconf
but it did not help.
That refactor from staging moved xorg-server into separate by-name so tracking what went where is quite hard. | 20:44:06 |
| ghpzin | I tried reordering these buildInputs like here on master:
https://github.com/NixOS/nixpkgs/blob/c30b80416b233c3673e573c0821bf96cc83818db/pkgs/servers/x11/xorg/overrides.nix#L452-L457
ie this diff on staging-next:
--- a/pkgs/by-name/xo/xorg-server/package.nix
+++ b/pkgs/by-name/xo/xorg-server/package.nix
@@ -139,10 +139,10 @@
xtrans
]
++ lib.optionals (!stdenv.hostPlatform.isDarwin) [
- dri-pkgconfig-stub
libdrm
libgbm
mesa-gl-headers
+ dri-pkgconfig-stub
]
++ lib.optionals stdenv.hostPlatform.isDarwin [
autoconf
but it did not help.
That refactor from staging moved xorg-server into separate by-name so tracking what went where is quite hard.
It is quite suspicious that upstream did something with glamor in 21.1.20 that broke other things, and now we get problems related to it too.
Maybe somewhere on staging between refactors somebody did something related to that update, but I could not find it so far. | 20:46:11 |
 Fabián Heredia | https://github.com/NixOS/nixpkgs/pull/472445
Fixes dee in staging-next (dependency of libunity <- Discord) | 22:36:32 |
| Fabián Heredia | * https://github.com/NixOS/nixpkgs/pull/472445
Fixes dee in staging-next (dependency of libunity <- Discord) due to gcc15 breakage | 22:36:48 |
| Fabián Heredia | error: Cannot build '/nix/store/w2vcan2incrnad11ysbj6xj4w509mf8v-dee-unstable-2017-06-16.drv'.
Reason: builder failed with exit code 2.
Output paths:
/nix/store/3ajwi6lggwj7pjzv6gyx0g095x4sidhp-dee-unstable-2017-06-16
/nix/store/7is6xw5sxqydjbfrph3c9ix5mg2rg4nj-dee-unstable-2017-06-16-dev
/nix/store/jvxyifr4yqqgpi5vgc0dc315kvdj5s4n-dee-unstable-2017-06-16-py
Last 25 log lines:
> | ^~~~~~~~
> In file included from /nix/store/j359ndglk11v5j47ra66lc1vxlk2dbjw-glib-2.86.3-dev/include/glib-2.0/glib.h:92:
> /nix/store/j359ndglk11v5j47ra66lc1vxlk2dbjw-glib-2.86.3-dev/include/glib-2.0/glib/gtimer.h:75:10: note: declared here
> 75 | gchar* g_time_val_to_iso8601 (GTimeVal *time_) G_GNUC_MALLOC;
> | ^~~~~~~~~~~~~~~~~~~~~
> dee-tool.c: In function 'dump_peer':
> dee-tool.c:242:31: error: passing argument 2 of 'g_timeout_add_seconds' from incompatible pointer type [-Wincompatible-pointer-types]
> 242 | g_timeout_add_seconds (1, _timeout_cb, NULL);
> | ^~~~~~~~~~~
> | |
> | gboolean (*)(void) {aka int (*)(void)}
> /nix/store/j359ndglk11v5j47ra66lc1vxlk2dbjw-glib-2.86.3-dev/include/glib-2.0/glib/gmain.h:949:54: note: expected 'GSourceFunc' {aka 'int (*)(void *)'} but argument is of type 'gboolean (*)(void)' {aka 'int (*)(void)'}
> 949 | GSourceFunc function,
> | ~~~~~~~~~~~~~~~~^~~~~~~~
> dee-tool.c:207:1: note: '_timeout_cb' declared here
> 207 | _timeout_cb ()
> | ^~~~~~~~~~~
> /nix/store/j359ndglk11v5j47ra66lc1vxlk2dbjw-glib-2.86.3-dev/include/glib-2.0/glib/gmain.h:200:20: note: 'GSourceFunc' declared here
> 200 | typedef gboolean (*GSourceFunc) (gpointer user_data);
> | ^~~~~~~~~~~
> make[2]: *** [Makefile:468: dee-tool.o] Error 1
> make[2]: Leaving directory '/build/dee/tools'
> make[1]: *** [Makefile:533: all-recursive] Error 1
> make[1]: Leaving directory '/build/dee'
> make: *** [Makefile:439: all] Error 2
For full logs, run:
nix log /nix/store/w2vcan2incrnad11ysbj6xj4w509mf8v-dee-unstable-2017-06-16.drv
error: Cannot build '/nix/store/gf57xfm523vm35j0xxr001fald6w6jr4-libunity-unstable-2021-02-01.drv'.
Reason: 1 dependency failed.
Output paths:
/nix/store/isx9q4c8dpbippw32vpjzgymxsw9yz2m-libunity-unstable-2021-02-01
/nix/store/vswssrm09dz6djws6a69hzmnj4spapcg-libunity-unstable-2021-02-01-dev
/nix/store/y706fvighhg56gmmwyr8ja5cvqjbhcs1-libunity-unstable-2021-02-01-py
error: Cannot build '/nix/store/2vzim0xvpgv17873p4wbw7n9gaxngac5-discord-0.0.119.drv'.
Reason: 1 dependency failed.
Output paths:
/nix/store/n4li8wz4rnn3jflrjgs30k3w1iz90hrx-discord-0.0.119
| 22:37:16 |
| ghpzin | So apparently this was load-bearing:
diff --git a/pkgs/by-name/xo/xorg-server/package.nix b/pkgs/by-name/xo/xorg-server/package.nix
index 271d279dd4..dea62e4973 100644
--- a/pkgs/by-name/xo/xorg-server/package.nix
+++ b/pkgs/by-name/xo/xorg-server/package.nix
@@ -151,6 +151,11 @@
mesa
];
+ hardeningDisable = [
+ "bindnow"
+ "relro"
+ ];
+
propagatedBuildInputs = [
dbus
libepoxy
and should not have been removed by refactor in https://github.com/NixOS/nixpkgs/pull/446052
Makes nixos tests with xserver work again.
Was there for 10 years before that: https://github.com/NixOS/nixpkgs/commit/27b9f5d65ee452c9a1a49a2e245b2a45ee1d65ab
Probably should have had comment with "Do not remove unless properly tested, without it errors during runtime"
Will send PR tomorrow. | 22:52:49 |
| K900 | H u h | 22:54:11 |
| K900 | Which one is it? | 22:54:15 |
| K900 | relro should definitely not matter at least | 22:54:31 |
| K900 | bindnow could be it ugh | 22:56:27 |
| K900 | I wonder why it needs that and xwayland does | 22:59:05 |
| K900 | Oh wait | 22:59:07 |
| K900 | It's meson isn't it | 22:59:09 |
| K900 | It's fucking meson | 22:59:13 |
| K900 | Because something something die GNU autotools | 22:59:41 |
| K900 | ...can we switch xorgserver to meson | 22:59:41 |
| ghpzin | Yes, rebuilt without "relro" tests pass.
So I guess will PR tomorrow with just "bindnow" and comment above to not remove without properly testing. | 23:01:04 |
| K900 | Does anyone want to attempt this lol | 23:01:35 |
| K900 | Or do we just let xorg users suffer | 23:01:35 |
