debug: update push hash mismatch doc — local verified, staging instrumentation deployed
sha256:38167901f0c4b5c714c56e85f444405ec51a358c3e5a9fea5f812245130cbdd1
sha
+22
~4
−28
symbols
sha256:28a4d7208bbf1570af3075548ed2356aaf6df58ad1bb7bbb02154ecd7b34c3f3
snapshot
+22
symbols added
~4
symbols modified
−28
symbols removed
0
dead code introduced
Semantic Changes
54 symbols
−
Level 0 — Unit: server correctly parses an O-frame
section Level 0 — Unit: server correctly parses an O-frame L90–100
−
Level 1 — Unit: delta+zlib reconstruction produces correct hash
section Level 1 — Unit: delta+zlib reconstruction produces correct hash L100–110
−
Level 2 — Integration: push a single delta object through stream path
section Level 2 — Integration: push a single delta object through stream path L110–118
−
Level 3 — End-to-end: muse push staging main with delta objects
section Level 3 — End-to-end: muse push staging main with delta objects L118–128
−
Root cause: field name mismatch between client and server
section Root cause: field name mismatch between client and server L50–86
−
The push used the stream path (not presign)
section The push used the stream path (not presign) L23–28
+
Current ceiling: what bytes does staging actually have for sha256:ab1054a8?
section Current ceiling: what bytes does staging actually have for sha256:ab1054a8? L102–139
+
Hypothesis B: R2 stored the object with different bytes due to a prior push bug
section Hypothesis B: R2 stored the object with different bytes due to a prior push bug L118–125
+
Hypothesis C: the Phase 2 push failed too (silently), or ab1054a8 was never stored
section Hypothesis C: the Phase 2 push failed too (silently), or ab1054a8 was never stored L125–131
+
Hypothesis D: decompress_if_needed called twice on the base
section Hypothesis D: decompress_if_needed called twice on the base L131–139
+
Fix 1: split_object_into_oc_frames field names (muse)
section Fix 1: split_object_into_oc_frames field names (muse) L162–171
+
Next ceiling: what bytes does staging actually have for sha256:ab1054a8?
section Next ceiling: what bytes does staging actually have for sha256:ab1054a8? L139–160
+
Presign path latent bug (ceiling 2, not yet triggered)
section Presign path latent bug (ceiling 2, not yet triggered) L181–193
+
Server is applying the delta (not treating as raw)
section Server is applying the delta (not treating as raw) L35–41
+
Server's apply_delta logic is correct (T5/T6 tests pass)
section Server's apply_delta logic is correct (T5/T6 tests pass) L60–65
+
The base object exists on staging (returns non-None)
section The base object exists on staging (returns non-None) L65–73
+
f97a3383 is not any obvious transform of known data
section f97a3383 is not any obvious transform of known data L73–84
~
Status
~
code@L4
Older
fix: use canonical MWP wire field names in push O-frames
sha256:131a9fc3542965349f2ee21922782a57d97e842418b432abc6976064b64426f1
All commits
Newer
debug: document root cause (presign path delta bytes) and repair log
sha256:4c140a69fa4995398308f3428caf10d39912d1d2b9fbf459d3094247143beed1
0 comments
To add a comment, use the Muse CLI:
muse hub commit comment sha256:38167901f0c4b5c714c56e85f444405ec51a358c3e5a9fea5f812245130cbdd1 --body "your comment"
No comments yet. Be the first to start the discussion.