mirror of
https://github.com/Klipper4FlashForge/firmware.git
synced 2026-10-22 05:12:16 -05:00
Oleksandr ShyshatskyiandClaude Opus 5
9b5b562a22
A sweep over every commented file in the tree. What is kept is what a reader needs to change the code safely: what a thing is, its signature, the invariants other files depend on, and the traps that fail silently (find's non-zero exit under set -e, s6's baked-in libexecdir, the -lm that decides whether _sqlite3 exists, opkg's compile-time prefix). What goes is the narration around it: incident histories, arguments against alternatives already rejected, restatements of the code below, and paragraphs re-explaining a rule three files already state. One paragraph in pkgs/3rdparty/opkg/build.sh was duplicated outright. Comment lines 7971 -> 6158. Every change is comment-only: no non-comment line differs anywhere in the 86 files, checked mechanically. Commented-out config directives (#fast_feed: 30000 and friends) are documentation of defaults and are kept. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NMMpQTajN7RDJySrhQ2cTE
48 lines
1.8 KiB
Bash
Executable File
48 lines
1.8 KiB
Bash
Executable File
#!/bin/sh
|
|
# The replica's stock entrypoint, with a writable /out inside the chroot.
|
|
#
|
|
# WHY A WRAPPER. /opt/printer/entrypoint.sh assembles the machine and then, as
|
|
# its last act, runs the case under `chroot /printer`. There is no hook
|
|
# afterwards, so a case that BUILDS something has no way to hand it back.
|
|
#
|
|
# AND WHY THE MOUNT IS NOT `docker run -v ...:/printer/out`. assemble.sh binds
|
|
# /printer-src onto /printer and remounts it read-only, and a bind does not
|
|
# carry the submounts underneath it -- a docker mount at /printer/out is
|
|
# buried the moment the machine is assembled, silently, leaving an empty
|
|
# directory. Mounting at /printer-src/out instead trips the `rm -rf
|
|
# /printer-src` assemble.sh opens with.
|
|
#
|
|
# So the mount belongs between assemble.sh and the case. Running the stock
|
|
# entrypoint with a case that does nothing makes that seam.
|
|
set -e
|
|
|
|
CASE="${1:?usage: entrypoint-out.sh <script-to-run-inside-chroot>}"
|
|
R=/printer
|
|
|
|
printf '#!/bin/sh\nexit 0\n' > /tmp/noop.sh
|
|
chmod +x /tmp/noop.sh
|
|
/opt/printer/entrypoint.sh /tmp/noop.sh
|
|
|
|
# The mountpoint is made through /printer-src, the writable side of the same
|
|
# bind: creating it under /printer directly would be a write to a read-only
|
|
# mount. The bind itself is fine over a read-only parent -- a mount is not a
|
|
# write to the filesystem it lands on, and it carries its own flags.
|
|
mkdir -p /printer-src/out
|
|
mount --bind /out $R/out
|
|
|
|
cp "$CASE" $R/tmp/case.sh
|
|
chmod +x $R/tmp/case.sh
|
|
set +e
|
|
chroot $R /bin/sh /tmp/case.sh
|
|
RC=$?
|
|
set -e
|
|
|
|
# Hand the results to the invoking user rather than to root. The replica runs
|
|
# privileged, so anything it writes to the bind mount lands root-owned in the
|
|
# repo -- and the build lane, which runs as the caller, then cannot delete its
|
|
# own scratch directory.
|
|
if [ -n "${OUT_UID:-}" ]; then
|
|
chown -R "$OUT_UID:${OUT_GID:-$OUT_UID}" /out 2>/dev/null || true
|
|
fi
|
|
exit $RC
|