--clean-build clears all caches (Go, NVIDIA, lb packages, work dir)
and rebuilds the Docker image, then proceeds with a full clean build.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Removes all cached build artifacts: Go cache, NVIDIA/NCCL/cuBLAS
downloads, lb package cache, and live-build work dir. Use before
a clean rebuild or when switching Debian/kernel versions.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Switch squashfs compression from xz to zstd (3-5x faster compression,
~10-15% larger but decompresses faster at boot)
- Stop rm -rf BUILD_WORK_DIR on each build; rsync only config changes
so lb can reuse its chroot across builds (skips apt install step)
- Keep lb-packages cache in CACHE_ROOT as fallback if work dir is wiped
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
apt-cacher-ng requires a separate container; lb's own package cache
persisted in --cache-dir is simpler and sufficient.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Saves cache/packages.chroot before wiping BUILD_WORK_DIR and
restores it after, so apt packages are not re-downloaded on every
build. Cache lives in --cache-dir (same place as Go/NVIDIA cache).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Pass APT_PROXY=http://host:3142 to build-in-container.sh to route
all apt traffic through a local cache. Also supports --apt-proxy flag.
Mirrors in auto/config are set from BEE_APT_PROXY env when present.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
toram was incorrectly added to the default bootappend-live causing
every boot to copy the full ISO to RAM (slow on BMC virtual media).
Default boot reads squashfs from media; toram is available as a
separate menu entry.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Removes ~100-300MB from the squashfs: man pages, non-en locales,
python cache, apt lists and package cache, temp files and logs.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Without this, git refuses to read the bind-mounted repo (UID mismatch)
and describe returns empty, causing the version to fall back to iso/v1.0.20.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Without toram the squashfs is read from the physical medium at runtime.
Disconnecting the USB/CD after boot causes SQUASHFS I/O errors on any
uncached block, making all X11 apps crash with SIGBUS.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Plain v2.x tags are now the active tagging scheme; iso/v1.0.x tags
are legacy. Swap priority in resolve_iso_version so the ISO is named
bee-debian12-v2.x-amd64.iso instead of v1.0.x-N-gHASH.
Also tighten the v* pattern to v[0-9]* to avoid accidentally matching
other prefixed tags in both resolve functions.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
resolve_iso_version only matched iso/v* pattern; GUI release tags
(v2, v2.1 ... v2.6) were ignored, falling back to the old v1.0.20
annotated tag via resolve_audit_version.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add video=1920x1080 to kernel cmdline (sets fbdev to Full HD)
- Update GRUB gfxmode to 1920x1080 (fallback to 1280x1024,auto)
- Add Xorg Monitor section with 1920x1080 Modeline and preferred mode
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
startx from user shell has /dev/fb0 permission issues and is fragile.
LightDM starts Xorg as root — standard LiveCD approach that works
on server hardware / IPMI KVM with nomodeset + fbdev/vesa.
- Add lightdm package, configure autologin as bee/openbox session
- Add /usr/share/xsessions/openbox.desktop
- Remove startx from .profile (LightDM manages X lifecycle)
- Remove Xwrapper.config needs_root_rights workaround (no longer needed)
- Enable lightdm.service in setup hook
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
startx from autologin shell targets VT1 directly — KVM sees the
graphical UI without VT switching. Remove bee-desktop.service
(systemd-launched X defaults to VT7, invisible on KVM).
Add xserver-xorg-video-fbdev for server AST/VGA framebuffer.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
grub-pc and grub-efi-amd64 conflict with each other in Debian 12.
The -bin packages provide the same grub-install binaries without conflict.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Copies the live system to a local disk via unsquashfs — no debootstrap,
no network required. Supports UEFI (GPT+EFI) and BIOS (MBR) layouts.
ISO:
- Add squashfs-tools, parted, grub-pc, grub-efi-amd64 to package list
- New overlay script bee-install: partitions, formats, unsquashfs,
writes fstab, runs grub-install+update-grub in chroot
Go TUI:
- Settings → Tools submenu (Install to disk, Check tools)
- Disk picker screen: lists non-USB, non-boot disks via lsblk
- Confirm screen warns about data loss
- Runs with live progress tail of /tmp/bee-install.log
- platform/install.go: ListInstallDisks, InstallToDisk, findLiveBootDevice
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Default boot no longer loads ISO to RAM (slow on BMC virtual media).
Separate menu entry added for toram in both GRUB and isolinux.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
nv/target has no .h suffix; use -type f instead of -name '*.h' to
detect non-empty include directories.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
cuda_fp16.h (included by cublas_api.h) requires <nv/target> from
the CUDA C++ Core Libraries (cuda-cccl-13-0).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
ls *.h missed headers in subdirectories like crt/host_defines.h;
use find -maxdepth 2 instead.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
cuda-crt-13-0 may not share the same version string as cuda-cudart-13-0;
pass empty version to lookup_pkg to match the first available version.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
cublasLt.h -> cublas_api.h -> driver_types.h -> crt/host_defines.h
which lives in the cuda-crt-13-0 package, not cudart-dev.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
NVIDIA CUDA .deb packages install headers under
/usr/local/cuda-X.Y/targets/x86_64-linux/include/ not /usr/include/,
causing copy_headers() to silently skip them.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
awk exit in the blank-line block jumps to END, which printed the
result again causing repo_sha to contain the hash twice with a newline,
breaking the sha256 string comparison.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Debian Packages.gz uses CRLF line endings; \r in the captured SHA256
field caused string comparison to fail even when hashes were identical.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Echo messages captured in stdout polluted the return value of
download_verified_pkg(), causing extract_deb() to receive a
multi-line string instead of a file path and silently exit via set -e.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
console=tty0 sent kernel messages to the active VT (tty1), overwriting
the TUI. Changed to console=tty2 so kernel logs land on a dedicated
console. tty1 is now clean; operator can press Alt+F2 to inspect kernel
messages and Alt+F3 for an extra shell.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
loglevel=3 was hiding all kernel messages on tty0/ttyS0 except errors.
Machine crashes (panics, driver oops, module failures) were silent on VGA.
Restored loglevel=7 so kernel messages up to debug are printed to both
tty0 (VGA) and ttyS0 (SOL). Journald MaxLevelConsole reduced to info
(was debug) to reduce noise on SOL while keeping it useful.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Previously the ISO file was named after git describe --match 'audit/v*',
so a new iso/ tag produced names like v1.0.9-1-gXXXXXXX instead of v1.0.17.
Now build.sh has resolve_iso_version() that looks at iso/v* tags separately.
The bee binary inside the ISO still uses AUDIT_VERSION_EFFECTIVE.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- build-nvidia-module.sh: copy libnvidia-ptxjitcompiler.so.* alongside
libcuda/libnvidia-ml — required by cuModuleLoadDataEx for PTX JIT.
Without it: CUDA_ERROR_JIT_COMPILER_NOT_FOUND at runtime.
Cache check updated to force rebuild when ptxjitcompiler is missing.
- bee-nvidia-load: run ldconfig after module load so that NVIDIA/NCCL
libs injected into /usr/lib/ are visible to dlopen() callers.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Commit d36e844 dropped console=tty0 and added dual-serial + debug logging.
Without console=tty0 the kernel never initialises the VGA console,
leaving the physical screen permanently blank.
- Restore console=tty0 (VGA) as primary, keep console=ttyS0 for SOL
- Drop console=ttyS1 (redundant second serial port)
- Replace loglevel=7 + journald debug flood with loglevel=3 (errors only)
so kernel messages don't overwrite the TUI on the local screen
- Remove systemd.log_target/forward_to_console debug params
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
AMD does not publish Debian Bookworm packages at all (only focal/jammy/noble).
Switch ROCM_UBUNTU_DIST to "jammy"; jammy packages install cleanly on
Debian 12 due to compatible glibc. Also expand candidate list to include
point-releases (6.3.4, 6.3.3, …) so we pick the latest actually-published one.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
