Repository navigation
setup: an interrupted engine update is tried again instead of skipped for good (#2159) - #2161
ByteJuggler wants to merge 1 commit into
Conversation
8fda2a5 to
f808495
Compare
…is tried again instead of skipped for good (Niko1221#2159) Source: Niko1221#2161 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…up robustness, conversation cache memory) Serve: Niko1221#2290 Niko1221#2284 Niko1221#2285 Niko1221#2170 Niko1221#2140 Niko1221#2278. Setup: Niko1221#2161 Niko1221#2287. Engine: Niko1221#2253 Niko1221#2265 Niko1221#2247 Niko1221#2217 Niko1221#2243. New opt-in engine flags: --session-state-codec (enabled in the local configs), --conversation-cache-park-gain and --drop-file-cache (left off: no measured benefit on the RX 7600 XT box). Bench: bench/results/upstream-prs-2026-10-13. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Checked on Linux: |
…ipped for good (Niko1221#2159) get_prebuilt moves BUILD.json aside before it downloads; when the process died before the failure branch put it back (a crash, a reboot, a closed window), every later update_installed_engine returned early on the missing stamp and the old engine stayed without a word. A leftover BUILD.json.prev beside the exe now goes back and the update runs again.
f808495 to
bc80631
Compare
|
Thanks for the pull request. We're resetting the tracker for 0.2.0-beta. There are more than 1,300 open issues and pull requests, too many for us to review properly, so we're closing everything that isn't being worked on right now. Closed here does not mean rejected. To reopen it takes about two minutes:
The rules are in CONTRIBUTING.md. |
Fixes #2159.
get_prebuiltmovesBUILD.jsontoBUILD.json.prevbefore downloading, andupdate_installed_engineonly puts it back if the process survives to itsnew is Nonebranch. After a crash, reboot or closed window mid-update, every later update returned early on the missingBUILD.json, so the install stayed on its old engine without saying so.Now, when
BUILD.jsonis missing butBUILD.json.prevand the exe are there, the stamp is restored (with a warning) and the update runs again.test_an_interrupted_update_is_tried_again(fails before the change, passes after).tools/suite: 3 failures intest_setup_engine_hash.WhatARefusalDoesToTheCaller, which fail the same way onmainwithout this change.