Update libcamera build instructions

[aallan]

Need to,

• The "lite" instructions need to replace "sudo pip3 install jinja2" by "sudo apt install python3-jinja2"

• We need to drop "sudo pip3 install pyyaml ply" for "sudo apt install python3-yaml python3-ply"

• We can get rid of the "sudo pip3 install --upgrade meson" - hurray!

• The libpisp repository wouldn't let me check it out, I assume the access permissions need changing!!

• I also notice the meson setup line still includes things like "-Dpipelines=raspberrypi -Dipas=raspberrypi". I thought we replaced "raspberrypi" ages ago by "rpi/vc4"?? Anyway, now it wants to be "-Dpipelines=rpi/vc4,rpi/pisp -Dipas=rpi/vc4,rpi/pisp" (or users who are building for only one of the platforms can obviously omit the other).

• libcamera-apps were unhappy because it claimed the sensor configuration was not a pointer (i.e. some confusion about the presence of std::optional). Maybe that will sort itself out once our packages all get updated? Possibly we can remove the "sudo ldconfig" too as that was a workaround for a cmake problem, and we now use meson/ninja?

[aallan]

See also #3325.

[JimJ92120]

hi @aallan not sure if relevant to add up here too (i'm also a bit new to Raspberry Pi eco-system)

following instructions may needs to be more descriptive: https://github.com/raspberrypi/documentation/blob/develop/documentation/asciidoc/computers/camera/rpicam_apps_building.adoc?plain=1#L88

# use -j 2 on Raspberry Pi 3 or earlier devices

I've encountered memory issue when running the build with ninja on Pi Zero 2 W . The process would appear to either get stuck or hang (or very very very slow).
Over 1h and it was still stuck on the same step. Retrying many times, with clean install as well didn't make it work.

Device: Raspberry Pi Zero 2 W
OS: Raspberry Pi OS Lite - Bullseye - 32bit version

---

Some research pointed out to possible RAM limitations, it would run out of resources when compiling.
This was also confirmed by running the same compilation on e.g laptop where it builds within 2-3 minutes.

Using the -j 2 flag lowers memory usage - very slow but it compiles (30+ minutes).
Otherwise, process would hang and never completes (would assume something may crash too at some point).

---

Suggested to maybe change that comment based on RAM available vs device version?

I'm not familiar / experienced enough to confirm but would assume, future devices (Pi 6, 7, Pi Zero 3, 4, etc) would encounter similar issue if compiler takes too much resources.

It may be possible to build on another (more powerful) machine and ship the build via ssh to the Pi - some symlinks created during the compilation would cause few issues though.

[JamesH65]

hi @aallan not sure if relevant to add up here too (i'm also a bit new to Raspberry Pi eco-system)

following instructions may needs to be more descriptive: https://github.com/raspberrypi/documentation/blob/develop/documentation/asciidoc/computers/camera/rpicam_apps_building.adoc?plain=1#L88

# use -j 2 on Raspberry Pi 3 or earlier devices

I've encountered memory issue when running the build with ninja on Pi Zero 2 W . The process would appear to either get stuck or hang (or very very very slow). Over 1h and it was still stuck on the same step. Retrying many times, with clean install as well didn't make it work.

Device: Raspberry Pi Zero 2 W OS: Raspberry Pi OS Lite - Bullseye - 32bit version

Some research pointed out to possible RAM limitations, it would run out of resources when compiling. This was also confirmed by running the same compilation on e.g laptop where it builds within 2-3 minutes.

Using the -j 2 flag lowers memory usage - very slow but it compiles (30+ minutes). Otherwise, process would hang and never completes (would assume something may crash too at some point).

Suggested to maybe change that comment based on RAM available vs device version?

I'm not familiar / experienced enough to confirm but would assume, future devices (Pi 6, 7, Pi Zero 3, 4, etc) would encounter similar issue if compiler takes too much resources.

It may be possible to build on another (more powerful) machine and ship the build via ssh to the Pi - some symlinks created during the compilation would cause few issues though.

use -j 1 (or remove the options entirely) on low-memory systems like the Zeros. They will quickly run out of memory when doing parallel builds and then will start swapping, which is very slow in comparison. This will be more noticeable on C++ builds as compiling things like templates is very RAM intensive.

[lurch]

OS: Raspberry Pi OS Lite - Bullseye - 32bit version

If you're doing stuff with libcamera then we'd recommend that you use Bookworm rather than Bullseye. And if you're running on a low-memory device like a Zero 2 W then you'll definitely want to stick with the Lite version!

[JimJ92120]

If you're doing stuff with libcamera then we'd recommend that you use Bookworm rather than Bullseye. And if you're running on a low-memory device like a Zero 2 W then you'll definitely want to stick with the Lite version!

unfortunately as of now, Bookworm isn't available for Pi Zero 2 W (v1 too) so not much an option
Lite version for sure

[JimJ92120]

use -j 1 (or remove the options entirely) on low-memory systems like the Zeros. They will quickly run out of memory when doing parallel builds and then will start swapping, which is very slow in comparison. This will be more noticeable on C++ builds as compiling things like templates is very RAM intensive.

good one for pointing out swap

---

with htop while compiling, about ~80% RAM usage on average (clean install, nothing running but that and htop):

-j 2:

-j 1:

[lurch]

unfortunately as of now, Bookworm isn't available for Pi Zero 2 W

It is, you just need to select "No filtering" in Raspberry Pi Imager, or download it directly from https://www.raspberrypi.com/software/operating-systems/

[JimJ92120]

@aallan forget my comments, i don't know how to read :man_facepalming:

https://github.com/raspberrypi/documentation/blob/develop/documentation/asciidoc/computers/camera/rpicam_apps_building.adoc?plain=1#L13

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.