Repository navigation
run-workbench: reach our anxnet.device, and check it is what was staged - #135
Merged
Merged
Conversation
AMINETXDUO_SANA2_VENDOR used ${:-1}, so an empty value could never
select anxnet.device; it is now ${-1}: unset means vendor, empty means
anxnet. The A2065 vendor-driver injection is gated on a non-empty
VENDOR so an anxnet run on that board does not stage Commodore's bytes.
An anxnet run exports DRIVER_NAME/DIR/DEVICE/CARD before _sana2_names,
as tools/classicwb.sh does, so the bytes are staged as anxnet.device and
not under the board's vendor file name.
The staged file's sha256 must equal the archive's anxnet.device whenever
sana2_select chose anxnet; the run fails closed otherwise.
The stale anxnet.device fixture no longer overwrites a staged anxnet
driver: that file stands in as the existing driver, and only the
"live file differs from the stale one" check is skipped for it.
The generic CARD_SELECTED match accepts the drawer-install form
AmiNetXDuo:Devs/Networks/<driver>.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The bridge lock skips SLIRP and run-workbench.sh took it about 660 lines below its HD/WORK wipe, so a second run of one tag in one checkout deleted the live guest's drive and then waited. tools/emu-rig-lock.sh gains rig_claim_drive/rig_drop_drive: an exclusive flock beside the bridge lock, keyed by the cksum of the drive's resolved path (the path is in the holder record), blocking up to AMINETXDUO_DRIVE_WAIT (1800 s) then 6, 2 when it cannot arbitrate. Two checkouts with one tag are two drives and stay parallel. Every launcher that starts the emulator claims its drive before its first wipe, restage, Xvfb, tcpdump or emulator start: amiberry-run.sh, run-workbench.sh, run-nfsmount.sh, run-smbmount.sh, run-console.sh, run-wbgrab.sh, and the standing classicwb.sh, demo-rtg.sh and console-instance.sh (wait 0: refused at once). One lock order everywhere, drive before bridge; the standing check never waits. Helpers that are not the guest drop the drive lock with the others. bridge-lock-selftest.sh adds drive_serial, drive_parallel, drive_refuse, lock_order_live (four contenders over two tags and two backends finish inside the bound) and launcher_order, which finds every script that starts the emulator and fails one with no drive claim, a claim after its first destructive line, or bridge before drive. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
rig_claim_drive takes the backend as its third argument. With no
flock(1) a non-bridged run gets one warning ("no flock: drive <path>
not locked; do not run two of this tag at once") and 0, unlocked; a
bridged one still exits 2. Every caller passes its backend
(run-wbgrab.sh passes none); run-workbench.sh claims the drive once
the backend is known, still before its first destructive line.
Standing launchers are unchanged: rig_standing_exempt still requires
flock(1) ahead of the drive claim.
bridge-lock-selftest.sh: noflock_drive, with flock(1) off PATH the
function returns 0 with the warning for slirp and 2 for ens18, and a
SLIRP amiberry-run.sh gets past the drive with the warning.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An anxnet run stages anxnet.device and no vendor file, so the Installer's driver-file fallback preselects nothing. The install boot had no board, so InstallNetProbe found nothing either: the card page came up with every box clear, installdrive pressed Proceed, the script answered "Select at least one network device." and the pair repeated until the timeout (ariadne2 drawer AVERAGE run at 24ade38, ~140 clicks, status 20). Boot the install with the board, as -M already does, so the probe preselects it. NO_CARD keeps its board-less boot. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
Comment on lines
+433
to
+435
| ( umask 000; : >> "$f" ) 2> /dev/null || { | ||
| echo "cannot create the drive lock $f" >&2; return 2; } | ||
| exec {fd}>>"$f" || { echo "cannot open the drive lock $f" >&2; return 2; } |
The drawer arm installs twice and marks S:Network-Startup between the runs; a boot-disabled install writes no S:Network-Startup, so the mark fails and the run ends at status 20 after a correct install. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
install/test/run-workbench.shcould not exercise our ownanxnet.device(re-implementation of codex's reviewed b23fc4de/3252c4c6, which could not be pushed from its responder):AMINETXDUO_SANA2_VENDOR="${…:-1}"mapped empty to 1, so anxnet was unreachable; now${…-1}(unset = vendor, empty = anxnet), comment corrected, and in the SAME commit the a2065 vendor-driver injection requires a non-empty VENDOR (else anxnet would be "selected" with vendor bytes).SANA2_SEL_SOURCE=anxnet, export the full set before_sana2_names(DRIVER_NAME=anxnet.device, DIR=Networks, DEVICE, CARD), astools/classicwb.shdoes; vendor runs export none, so they keep their own names.AmiNetXDuo:Devs/Networks/$SANA2_DRIVER(the installer: name a drawer install's drivers through AmiNetXDuo: #127 drawer form), no glob.SANA2_SEL_SOURCE=anxnet: staged sha256 must equal the archive'sDevs/Networks/anxnet.device; fails closed.anxnet.devicefixture no longer overwrites a staged anxnet driver; backup checks still apply.Selection proof (the real block +
tools/sana2-stage.sh, fixture drivers): empty VENDOR →source=anxnet staged=DEVS:Networks/anxnet.device bytes=OUR-anxnetfor ne2000_pcmcia, ariadne2, a2065; unset/0 → vendor bytes and names, nothing exported. origin/main:source=vendorfor all empty rows.check-harnessesPASS, shellcheck 0.End-to-end drawer run (
AMINETXDUO_SANA2_VENDOR= … -D -l AVERAGE -N ariadne2 -B) follows on the rig.🤖 Generated with Claude Code