+ Wj9Rt^RIHt^RIt^RIt^RIHt^RIHtRRlt Rt RR lt R R lt R R lt RRltRRltRRltRRltRRltRRltRRltRRltRRltR R!ltR"R#ltR#)$a{Steps of ``hermes update --check``: debris cleanup, channel target, scoped fetch, verdict. ``update_cmd._cmd_update_check`` (a frozen updater surface, see ``tests/compat``) orchestrates these. Facade helpers are read through ``_uc()`` at call time and origin helpers are imported per function, so test patches on ``hermes_cli.update_cmd`` and the origin modules stay effective. ) annotationsN)Path)Anyc ,V^8dQhRRRRRRRRRR /#) git_cmd list[str]rootrargskwargsrreturnz subprocess.CompletedProcess[str])formats"OC:\Users\johnn\AppData\Local\hermes\hermes-agent\hermes_cli\update_cmd_check.py __annotate__rs0  ) 4 y C Ll c ^RIHpVPR^4V!4,VR&^RIHpV!W3RVRRRRRR R R /VB#) )windows_hide_flags creationflags)run_gitcwdcapture_outputTtextencodingzutf-8errorsreplace)hermes_cli._subprocess_compatrgethermes_cli.update_custodyr)rr r r rrs&&&, r_gitr sh@%jj!>i>t>S>c>d>rcaaaaaa a ^RIHp^RIHoHo ^RIHo^RIHo RVVV V3RllpRVV V3RllpV!SW%VR 7#) zc`hermes update` would refuse ``target`` (see update_downgrade_guard), so --check must not offer it.) would_refuse)NO_LAZY_FETCH_ENVnoninteractive_git_env)_github_predates_runtime)get_hermes_homec V^8dQhRRRR/#)rr rBr intr )rs"rr_refuses..__annotate__Hs33rc<\SS\V4/S!4CSC^R7P# \\P 3dRu#i;i))envtimeout)r list returncodeOSError subprocessSubprocessError)r rGrrHr s*rgit_refuses..gitHsX tDz7h:P:R7hVg7h "$$.J /334 I s+/AAcV^8dQhRR/#)rsharBr )rs"rrrMOswwCwrc><S!SVS!4R, R, 4#)z source-checkszrelease-runtime.jsonr )r[rIrJrCs&rremote_refuses..remoteOs ' C9J_9\_u9uvvr)rXr]) !hermes_cli.update_downgrade_guardrFrrGrHhermes_cli.source_checkrIhermes_constantsrJ) rr rArCrFrXr]rGrIrJrHs ff&f @@@@r_refusesrbAs4>W@0ww ff ==rc(V^8dQhRRRRRRRR/#) rselected_channelrBrrr rr str | Noner )rs"rrrTs)&&S&9&D&U_&rc<^RIHpHp^RIHp^RIHpHp\RV 24V!V!V!44V4pV!WV4p V!WW)R7p X P'g V P #V P"'d\R V R V P$ 24V P&'d\R V P( R 24R#\+4P-W4V P8Xd\R V P( R24R#\/WV PV P04'd'\RV P( R24\R4R#\3WRRV PR.4P4^8Xd'\RV P( R 24\R4R#\RV P( 24\R4R# \\\P3d3p \RT RT 24\P!^4Rp ? ELRp ? ii;i)aReport a release-pinned channel's verdict, or return the branch to compare against. ``None`` means the verdict is printed (the channel pins a commit); exits 1 when the channel cannot be resolved. An install riding its default channel only ever moves forward. )get_config_path$require_readable_config_before_writeresolve_source_targetchannel_recordrides_default_channel→ Update channel:  forward_only✗ Could not resolve the  source channel: N→  retired; source destination: u✓ No newer release: .u(✓ Up to date with the latest release ().u→ Holding at this build: z0 predates the gateway runtime your sessions use.z `hermes update` stays on this build; the next release that includes the runtime installs normally (`hermes update --set-channel main` follows every commit).z merge-basez --is-ancestorHEADu6→ This checkout is newer than the selected release: zj `hermes update` moves it back to that release (`hermes update --set-channel main` follows every commit).u → Selected release available: z$ Run `hermes update` to install it.)hermes_cli.configrgrhhermes_cli.source_releasesrjhermes_cli.update_channelrlrmr/rU ValueErrorrVrWsysexitcommitbranchretiredchannelaheadlabelr%_capture_head_sharbrCr rT) rdrr rgrhrjrlrmrecordrprAexcs &&& rchannel_compare_branchrTs X@O !1 2 34 @ARSUY ZF(4HL&'7$b ===}} ~~~ %&&DV^^DTUV ||| &v||nA67    /6== @ 8bIJ  '0A0A B B +FLL>9ijk ^ _  glOV]]FS T _ _cd d Fv||nTUVW L M  0?@ 45 / Z!;!; < *+;*</``, whatever ``remote..fetch`` says. Narrow clones (tag-pinned ``--single-branch``, #125112) map only the tag, so a by-name fetch writes FETCH_HEAD and never the tracking ref the updater resolves. The ``+`` is load-bearing: on a depth-1 clone the new tip is not a descendant of the old one. z +refs/heads/z:refs/remotes//r )r]rs&&rtracking_refspecrs&xq AArc ,V^8dQhRRRRRRRRRR/#) rrrr r depth_argsr]rBrr )rs"rrrs0))I)T)y)#)WZ)rc \RV R24\WR.VOVN\W44N3/\4P 4B#)u→ Fetching from z...fetch)r/r rr%_no_prompt_git_kwargs)rr rr]rs&&&&&r_fetchrsQ vhc *+ W*WfW6Fv6VW ) % % % ' ))rc(V^8dQhRRRRRRRR/#)rrrr rrrBrr )rs"rrrs/qq)q4qqR[qrc aVR8XdF\VS.R O4P^8Xd(\VSVRV4pVP^8XdVRV 23#^RIHp\ R4V!V3RlVR.VORN\ RV4NS4R V 23#) aCFetch only ``branch`` and return ``(fetch_result, compare_ref)``. A bare ``git fetch `` pulls every ref, and this repo has thousands of auto-generated branches. ``main`` prefers upstream as the canonical reference; other branches go straight to origin, because a fork's branch usually has no upstream counterpart. mainupstreamz upstream/)!fetch_with_partial_clone_recoveryu→ Fetching from origin...cL<\VSV3/\4P4B#)N)r r%r)gcar s&&r&fetch_compare_branch..sd2tQH#%*E*E*GHrroriginorigin/)r]zget-urlr)r rTrr.rr/r)rr rr fetch_resultrs&f&& rfetch_compare_branchrs @ A L LPQ Q!'4ZPL&&!+#y%999D '( ,H'UJUU2B8V2TUW[ ]`ghngo^p qqrc V^8dQhRRRR/#r'r )rs"rrrs Y Y Y Yrc^RIHpHpV!V4pV'd\RV R24V!V4pV'd\RV R24R#R#)zDrop the stale ``.git/shallow`` grafts a depth-1 fetch leaves behind. Git never removes old grafts; unpruned, the file keeps growing and merge-base / the orphan-divergence heuristic stop working (#105951). )prune_stale_shallow_grafts repair_broken_shallow_boundariesz (restored z broken shallow boundary(ies))z (pruned z4 stale shallow graft(s) left by past depth-1 checks)N)r.rrr/)r rrrepairedpruneds& rrepair_shallow_graftsrsK `/5H XJ&DEF ' -F  6("VWXrc(V^8dQhRRRRRRRR/#) rrrr rcompare_branchrBr rDr )rs"rrrs.ee eeseterc<\WRRRV.4P^8H#)rz--verifyz--quiet)r rT)rr rs&&&rcompare_ref_existsrs$  ZNS T _ _cd ddrc(V^8dQhRRRRRRRR/# rrrr rrrBr r(r )rs"rrrs/AAIATA3ASWArc \WRR.4PP4p\WRV.4PP4pV'dV'dW48Xd\R4R#^RIHp^RIHpV!W44pV^8Xd\R4R#Ve"V^8XdRMRp\R V R V R V R 24M\R V R24\RV!4 R24R#)zReport behind-ness without local history across the shallow boundary. Compares tip SHAs (mirrors the banner's ``_check_via_local_git``), then asks the GitHub compare API for the exact count: the remote graph is complete even when the local one is not. rrw✓ Already up to date.Nrecommended_update_command)_github_compare_behindr~commits⚕ Update available:   behind ruu⚕ Update available (behind rv Run ' ' to install.)r rrr/rxrr`r) rr rhead_sha target_sharrcounted commits_words &&& rreport_shallow_verdictrs GK#89@@FFHHgk>%BCJJPPRJJ8#9 '(<>$X:G!| '(#*a:J2KY%W_cdJ ""((* +F { '(%{8 L "6(!L>.AQQR ST< G.01 ?@rc$V^8dQhRRRRRR/#)rrrBrequestdictr tupler )rs"rrrs&)E)Ec)ED)Eu)Erc ^RIHp^RIHp^RIHp^RIHp^RIH p H p ^RI H p \VRR4'd R V 2RR R3#\4P4P p \4P#V4p V!V !V!\%VR ,4R , 4V 44p\'R V 24V!4;_uu_4V!WV V !WV 4R7pRRR4XP4'dY\'RT RTP6 24\TRR4'g+TP9RR4T 8XdRTRTP6/TR&TP:'dm\'RTP<'dRMR RTP> 24TP:TR&TP:TP:TP<TP@3#TPBfQhTPBTR&R TPB 2RR TP@3# +'giEL5;i \(\*\,P.3dCp\'RT RT R24T!4T !R4\0P2!^4Rp?ELRp?ii;i)aXResolve the update's target before any tree write: ``(target_ref, release_sha, target_is_head, repository)``; records branch/expected_sha/retirement on ``request``. ``target_is_head`` means an unchosen default subscription is already past the release, so HEAD itself is the target. Exits 1 (after ``stop()``) when the channel is unresolvable. )deepcopy)rh)retrying_readsrirk) _record_stoprNrFhomez config.yamlrnrorqrrz. No update was applied.channel_unresolvedrsrtrroriginal destinationchannel_retirementReleasezLatest releasez: expected_sha)"copyrrxrhhermes_cli.release_channelsrryrjrzrlrmhermes_cli.update_cmd_commonrgetattrr%_m PROJECT_ROOT_update_run_channelrr/rUr{rVrWr|r}rrrr~rrrCr)r rrrstoprrhrrjrlrmrr selectedrrArs&&&$$ rselect_apply_targetrs'F:@O9tXt$$!444 588: " "Du((.H'K WV_ -(/0467H  +,   *8d8Mhbf8giF~~~ XJrsv#   +$ l>&&RbB)q* Y e A6 A)EXr