@smart-cloud/publisher-exporter 1.2.3 → 1.2.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -41,6 +41,52 @@ manifests. The declaration does not require longer render readiness waits.
41
41
 
42
42
  ## Release notes
43
43
 
44
+ ### 1.2.4
45
+
46
+ - Restrict retry crawls to their initial failure URL list. Links discovered by
47
+ the retried pages remain useful manifest metadata, but cannot queue successful
48
+ old pages or newly discovered pages. This repair scope ignores changes to
49
+ unrelated pages and does not become a new incremental crawl.
50
+ - Reuse previously completed assets by default for retry and URL crawls, without
51
+ fetching their current change tokens or rewriting cached text assets again.
52
+ Local files must exist within the output tree; workspace objects must pass
53
+ scoped metadata, size, checksum, ETag and version checks. Missing, new, empty
54
+ or pending-rewrite assets are fetched normally. Current-job profile-bound
55
+ configuration/licence assets retain their mandatory fresh capture policy.
56
+ Use `--refresh-assets` to request fresh export downloads explicitly.
57
+ - Browser rendering still requests the origin resources it needs. This cache
58
+ policy avoids repeated exporter/API asset downloads; rewritten export files
59
+ are not injected into the source browser as if they were original assets.
60
+ - Observe document readiness through Playwright's lifecycle and isolated DOM
61
+ selectors, within the existing readiness budget. Same-document history URL
62
+ changes and page scripts replacing `window.URL` no longer cause false timeouts.
63
+ Diagnostics distinguish HTTP response arrival, completed response transfer,
64
+ document commit, observed lifecycle and unavailable DOM snapshots.
65
+ - Treat page-level deploy rewrite-guard rejections as partial publication:
66
+ upload safe files, retain the repair plan and failure ledger, and suppress all
67
+ deletions. Pending HTML is never uploaded. Apply the guard to SDK, Lambda,
68
+ S3-native and profile-prepared outputs. Actual upload failures, pending text
69
+ assets and strict content-sync transactions retain their failure barriers.
70
+ - Recheck partial-deploy state before establishing a normal-release baseline,
71
+ including when a guard rejection was first discovered during deploy.
72
+ - Accumulate undeployed output changes across successive normal, URL and retry
73
+ crawls. Six URL jobs followed by one deploy upload all six pending outputs;
74
+ a later job no longer replaces the earlier jobs' upload list. The incomplete
75
+ crawl marker still blocks deployment during interrupted work. Sequence-bound
76
+ content-sync plans remain separate transactions.
77
+ - Preserve a canonical page's file when removing a stale query alias sharing
78
+ the same path. Query variants cannot overwrite a known canonical output;
79
+ unresolved collisions are recorded for repair instead. No query parameters
80
+ are discarded globally, since other sites may use them for distinct content.
81
+ - Distinguish manifest-wide unfinished rewrite entries from the files actually
82
+ blocked in the current deploy in guard diagnostics.
83
+
84
+ Upgrade the coordinator and rebuild render-worker images with 1.2.4 to receive
85
+ the shared readiness correction. No WordPress plugin update or browser reinstall
86
+ is required. Targeted asset reuse does not replace normal full/incremental
87
+ publication when source assets intentionally change. For a repair with fresh
88
+ assets, run `publisher-exporter crawl --retry-failed --refresh-assets`.
89
+
44
90
  ### 1.2.3
45
91
 
46
92
  - Continue normal full, incremental and URL crawls when individual pages fail