Openindiana 2025 issues

[lucho]

Replaced Solaris 11.4 whose updates are extremely difficult to get with OpenIndiana 2024 and updated it on 31 March 2025. Installed Mate. Four problems arised:

1. Black screen during bootstrap until X starts.
2. Firefox can't play certain videos (e.g. http://youtu.be/8uMRa60nhKg).
3. Today, after "sudo pkg update", I got "pkg: unable to activate openindiana-2025:04:02". *
4. Sites using "infinite scrolling" may make Firefox allocate memory until the whole OS crashes.

*) The "sudo beadm activate -v openindiana-2025:04:02" command responds with the following:

be_do_installboot_walk: child 0 of 1 device c6t3d0s4
be_do_installboot_helper: install failed for device c6t3d0s4.
  Command: "/usr/sbin/installboot  -b /tmp/tmpgu_2dfn3/boot /dev/rdsk/c6t3d0s4"
  Errors:
Error reading bootblock from /tmp/ibootWLW3ea/EFI/Boot/bootia32.efi
Bootblock version installed on /dev/rdsk/c6t3d0s4 is more recent or identical to
/tmp/tmpgu_2dfn3/boot/gptzfsboot
Use -F to override or install without the -u option.

Bootblock version installed on /dev/rdsk/c6t3d0s4 is more recent or identical to
/tmp/tmpgu_2dfn3/boot/gptzfsboot
Use -F to override or install without the -u option.

sh: 1309: Memory fault
be_run_cmd: command terminated on signal: Segmentation Fault
Unable to activate openindiana-2025:04:02.
Error installing boot files.

[Onauk]

Originally Posted by lucho

1. Firefox can't play certain videos (e.g. http://youtu.be/8uMRa60nhKg).

This seems like a non available codec. This video has only been encoded with AVC1 and support in Firefox is available with version 100 onwards.

You can know the encoding either by right clicking on the video and choosing "Stats for nerds" or by using `yt-dlp -F video-url`.

[lucho]

Thanks! I couldn't see the difference between the parameters of playable and non-playable videos with "yt-dlp -F", but the latter are fewer than the former, so that's not a big problem. Firefox version is 136.0.4.

[Onauk]

There seems to be a solution to this AVC1 issue in Firefox shared on the OpenIndiana mailing list:

https://www.mail-archive.com/openind.../msg26424.html

I have not tested it!

[lucho]

Thank you very much for this hint! I added the "hipster-encumbered" repository (publisher) to the main one ("openindiana.org") as explained there, and then installed the "dav1d" AND "ffmpeg" packages from there. Et... voilà! The previously non-playable videos are now playable in Firefox!!!

So now I'm going to strike out the second issue in my list.

[lucho]

I think I've managed to resolve the third issue too. It turned out that the problem was that I've installed Openindiana in the former Solaris partition and the EFI partition on the disk was so small that there is no room there for even one of the EFI boot files. Truncated both EFI files ("loader32.efi" and "loader64.efi") to 1536 bytes each (there is no room for any of them on the EFI partition anyway) and this time the activation command "sudo beadm activate -v openindiana-2025:03:31" was successful:

be_do_installboot_walk: child 0 of 1 device c6t3d0s4
  Command: "/usr/sbin/installboot  -b //boot /dev/rdsk/c6t3d0s4"
  Output:


bootblock written to /tmp/iboot9VWHfa/EFI/Boot/bootx64.efi

bootblock written to /tmp/iboot9VWHfa/EFI/Boot/bootia32.efi

//boot/pmbr is newer than one in /dev/rdsk/c6t3d0p0
stage1 written to slice 0 sector 0 (abs 0)

  Errors:
Error reading bootblock from /tmp/iboot9VWHfa/EFI/Boot/bootia32.efi
Error reading bootblock from /tmp/iboot9VWHfa/EFI/Boot/bootx64.efi
Bootblock version installed on /dev/rdsk/c6t3d0s4 is more recent or identical to
//boot/gptzfsboot
Use -F to override or install without the -u option.

Bootblock version installed on /dev/rdsk/c6t3d0s4 is more recent or identical to
//boot/gptzfsboot
Use -F to override or install without the -u option.

Activated successfully

The first problem turned out to be an unknown console issue. Finally managed to compose the following menu for Linux's GRUB that shows the boot menu and kernel boot messages while booting at last:

menuentry 'OpenIndiana (by /etc/grub.d/39_custom)' {
	echo Loading...
	insmod part_gpt
	insmod zfsinfo
	search -s -l rpool1
	kfreebsd /ROOT/openindiana-2025:03:31@/boot/loader
}

Note that the "echo" command is required for the screen output to work during bootstrap. It "uncorks" it somehow...

[Onauk]

Congrats on solving all the issues!

[lucho]

Thanks! Actually, there is a Firefox bug I discovered recently which I added to the list in my first message, and some minor problems left like the missing "i965_dri.so" file in the "x11/library/mesa" package (here's a line from my "/var/log/Xorg.0.log"):

(EE) AIGLX error: dlopen of /usr/lib/xorg/modules/dri/amd64/i965_dri.so failed (ld.so.1: Xorg: fatal: /usr/lib/xorg/modules/dri/amd64/i965_dri.so: open failed: No such file or directory)

but still the video playback is fast enough, except when the picture is scaled.

For now, I've switched to another task, on which I'm going to post elsewhere later.

[Onauk]

Mesa package should ship with i965 if it is supported by the OS but I don't see any mention of i965 in OpenIndiana documentation.

The file list of x11/library/mesa doesn't include i965[1].

You could look to the components page to see if your hardware is listed as supported by i915[2] or you could use the graphics stack page[3] for more details.

[1]: https://pkg.openindiana.org/hipster/...250118T232949Z
[2]: https://docs.openindiana.org/communi...phics-adapters
[3]: https://docs.openindiana.org/dev/graphics-stack/

[lucho]

My "Intel HD Graphics 4600" (i7-4790 "Haswell") is supported. Probably "i965_dri.so" doesn't build for some reason. I can try to build it from the ports tree later.

[lucho]

Built "i965_dri.so" the following way. First, installed the necessary development packages:

sudo pkg set-publisher --non-sticky openindiana.org
sudo pkg change-facet facet.require.consolidation/userland/userland-incorporation=false
sudo pkg uninstall userland-incorporation
sudo pkg install build-essential
sudo pkg install gcc-14
git clone https://github.com/OpenIndiana/oi-userland.git

(The first 3 of the above commands are necessary to install and update later if necessary the otherwise forbidden "cmake" package, which is one of the many packages installed when we install "build-essential".)

Made the following change:

--- oi-userland/components/x11/mesa/Makefile-   вт апр 15 10:23:17 2025
+++ oi-userland/components/x11/mesa/Makefile    вт апр 15 12:21:17 2025
@@ -65,7 +65,7 @@
 CONFIGURE_OPTIONS += -Dshared-glapi=enabled
 CONFIGURE_OPTIONS += -Dgallium-xvmc=enabled
 CONFIGURE_OPTIONS += -Dgallium-xa=enabled
-CONFIGURE_OPTIONS += -Ddri-drivers=''
+CONFIGURE_OPTIONS += -Ddri-drivers='i965'
 CONFIGURE_OPTIONS += -Ddri-drivers-path='$(X11_SERVERMODS_DIR)/dri$(SERVERMOD_SUBDIR)'
 CONFIGURE_OPTIONS += -Delf-tls=false

Built it:

cd oi-userland-oi-hipster/
gmake setup
cd components/x11/mesa
gmake publish

Put the so built file in place:

sudo cp -p build/prototype/i386/usr/lib/xorg/modules/dri/amd64/i965_dri.so /usr/lib/xorg/modules/dri/amd64/

and performed a reboot. The result in "/var/log/Xorg.0.org" was:

[   243.649] (EE) Backtrace:
[   243.649] (EE) 0: /usr/bin/Xorg (xorg_backtrace+0xb1) [0x5062ab]
[   243.649] (EE) 1: /usr/bin/Xorg (OsSigHandler+0x73) [0x50ec80]
[   243.649] (EE) 2: /lib/amd64/libc.so.1 (__sighndlr+0x6) [0x7fffaf3c6f76]
[   243.649] (EE) 3: /lib/amd64/libc.so.1 (call_user_handler+0x1d1) [0x7fffaf3b98f1]
[   243.649] (EE) 4: /lib/amd64/libc.so.1 (sigacthandler+0xee) [0x7fffaf3b9bee]
[   243.649] (EE) 5: ?? [0xffffffffffffffff]
[   243.649] (EE) 6: /usr/lib/xorg/modules/dri/amd64/i965_dri.so (brw_disk_cache_init+0x78) [0x7fffa8b9bad8]
[   243.649] (EE) 7: /usr/lib/xorg/modules/dri/amd64/i965_dri.so (brw_init_screen+0x837) [0x7fffa8bdb4e7]
[   243.649] (EE) 8: /usr/lib/xorg/modules/dri/amd64/i965_dri.so (driCreateNewScreen2+0x1fd) [0x7fffa8be0edd]
[   243.649] (EE) 9: /usr/lib/mesa/modules/extensions/amd64/libglx.so (__glXDRIscreenProbe+0x198) [0x7fffaaac5b3a]
[   243.649] (EE) 10: /usr/lib/mesa/modules/extensions/amd64/libglx.so (GlxExtensionInit+0x147) [0x7fffaaab6b70]
[   243.649] (EE) 11: /usr/bin/Xorg (InitExtensions+0x6e) [0x56f00c]
[   243.649] (EE) 12: /usr/bin/Xorg (dix_main+0x303) [0x4c2324]
[   243.649] (EE) 13: /usr/bin/Xorg (main+0x28) [0x51dedf]
[   243.650] (EE) 14: /usr/bin/Xorg (_start_crt+0x87) [0x4a4417]
[   243.650] (EE) 15: /usr/bin/Xorg (_start+0x18) [0x4a4378]
[   243.650] (EE) 
[   243.650] (EE) Segmentation Fault at address 0x10
[   243.650] (EE) 
Fatal server error:
[   243.650] (EE) Caught signal 11 (Segmentation Fault). Server aborting

So, that's why the file "i965_dri.so" is missing from the "mesa" package.

[lucho]

The following line of the above log says that the crash was in the "brw_disk_cache_init" function:

[ 243.649] (EE) 6: /usr/lib/xorg/modules/dri/amd64/i965_dri.so (brw_disk_cache_init+0x78) [0x7fffa8b9bad8]

So I decided to disable the shader disk cache:

--- oi-userland/components/x11/mesa/Makefile-   вт апр 15 10:23:17 2025
+++ oi-userland/components/x11/mesa/Makefile    ср апр 16 10:03:32 2025
@@ -65,7 +65,8 @@
 CONFIGURE_OPTIONS += -Dshared-glapi=enabled
 CONFIGURE_OPTIONS += -Dgallium-xvmc=enabled
 CONFIGURE_OPTIONS += -Dgallium-xa=enabled
-CONFIGURE_OPTIONS += -Ddri-drivers=''
+CONFIGURE_OPTIONS += -Ddri-drivers='i965'
+CONFIGURE_OPTIONS += -Dshader-cache=false
 CONFIGURE_OPTIONS += -Ddri-drivers-path='$(X11_SERVERMODS_DIR)/dri$(SERVERMOD_SUBDIR)'
 CONFIGURE_OPTIONS += -Delf-tls=false

Rebuilt the driver and copied it to its place again (see above how). Voilà! It started to work:

[    98.167] (II) AIGLX: enabled GLX_MESA_copy_sub_buffer
[    98.167] (II) AIGLX: enabled GLX_ARB_create_context
[    98.168] (II) AIGLX: enabled GLX_ARB_create_context_profile
[    98.168] (II) AIGLX: enabled GLX_EXT_create_context_es{,2}_profile
[    98.168] (II) AIGLX: enabled GLX_INTEL_swap_event
[    98.168] (II) AIGLX: enabled GLX_SGI_swap_control
[    98.168] (II) AIGLX: enabled GLX_EXT_framebuffer_sRGB
[    98.168] (II) AIGLX: enabled GLX_ARB_fbconfig_float
[    98.168] (II) AIGLX: enabled GLX_EXT_fbconfig_packed_float
[    98.169] (II) AIGLX: GLX_EXT_texture_from_pixmap backed by buffer objects
[    98.172] (II) AIGLX: Loaded and initialized i965
[    98.172] (II) GLX: Initialized DRI2 GL provider for screen 0

Can't attach the so built "/usr/lib/xorg/modules/dri/amd64/i965_dri.so" file – even as a "tar.bz2" file, it's 7 MB – much larger than the 2,5 MB limit. So I uploaded the file here in a ZIP file.

[Onauk]

Nice work! This will be helpful to others for sure. Maybe you could add this info on the OpenIndiana wiki? That would be top notch

[lucho]

Where is this wiki? I couldn't find one. Wouldn't it be better to submit the aforementioned Makefile patch? Or even better – find the exact bug in the "brw_disk_cache_init" function in the brw_disk_cache.c file? But alas, I would hardly be able to do so, however short this function is.

[Onauk]

I was thinking about docs.openindiana.org but you are right, it's not a wiki and the patch you submitted is a way better solution.

If I may, to answer your question on github, to do a PR, you must first fork the repo, create a new branch on this repo then commit your patch on this branch. After this you will be able to do a pull request which will merge your changes from the branch of your fork to the official openindiana repo.

Best of luck!

[lucho]

Thanks for your explanation! I've worked with Subversion in the past, but Git is more complex and I must change my way of thinking to use it, which is easier to say than to do.

In any case, you helped me understood why it's called a "pull request" and not a "push request". They will "pull" the change from my forked repository, instead of letting me "push" it myself. So they don't need to grant me write access to the main repository, as in Subversion.

[Onauk]

Glad it was helpful, you should know that this is a github specific way of doing things (which was adopted by other forges like gitlab) but not everyone does it this way.

A well known example is Linux kernel development which only accept patches by email and then a Linux developer with write access will do the commit himself.

[lucho]

In both cases, the difference with the earlier systems like SCCS, CVS and SVN is that there are two levels of trust – trusted developers who are eligible to commit changes to the repository and non-trusted ones who can only propose changes but may not commit them.

Anyway, it was the necessity of changing the way of thinking that didn't let me understand the C++ language whereas I do understand C. Let's hope that this won't repeat this time.