[pjbgf]

https://entire.io/gh/entireio/cli/trails/1598

The helper printed its whole Go error chain after "fatal:". Users saw protocol steps (stateless-connect, info/refs, the advertisement fetch), replica failover details, dial errors and full node URLs. None of it says what went wrong or what to do next. The same 404 also read differently depending on the git protocol version.

Known failures now print one line about the cluster and the repository. The full chain still goes to the ENTIRE_DEBUG trace. The transport returns typed errors (HTTPStatusError, ErrNoNodeSucceeded), so the top level picks the message by type instead of matching strings.

Repository not found:

  before: fatal: stateless-connect v2 info/refs: repository not found:
          https://<node>/gh/o/r
  before: fatal: list git-upload-pack info/refs: fetch git-upload-pack
          advertisement: repository not found: https://<node>/gh/o/r
  after:  fatal: repository gh/o/r not found on <cluster>

No access (clone or push):

  before: fatal: list git-receive-pack info/refs: fetch
          git-receive-pack advertisement: server error (HTTP 403):
          permission denied
  after:  fatal: you do not have access to gh/o/r on <cluster>
          (permission denied)

Login rejected:

  before: fatal: stateless-connect v2 info/refs: authentication failed
          even after refreshing credentials - run 'entire login' if
          this persists
  after:  fatal: <cluster> rejected your login. Run `entire login`,
          then try again

Server error:

  before: fatal: stateless-connect v2 info/refs: server error
          (HTTP 503): db timeout
  after:  fatal: <cluster> could not handle the request
          (HTTP 503: db timeout). Try again later

Every replica unreachable:

  before: fatal: stateless-connect v2 info/refs: fetching info/refs
          after entry-domain redirect: all 2 attempted nodes failed,
          last error: Get "https://<node>/gh/o/r/info/refs?service=
          git-upload-pack": dial tcp <ip>: connect: connection refused
  after:  fatal: could not reach any server for gh/o/r on <cluster>.
          Check your connection and try again

Other noise removed:

• The replica failover path no longer puts HTML error pages from load balancers into the error. The entry URL path already dropped them.
• An unrecognised GIT_PROTOCOL version is now noted in the debug trace instead of printing a warning on stderr.
• Ctrl-C prints "interrupted" instead of a wrapped context error.

Redact URLs:

  before: fatal: invalid URL "entire://alice:s3cret@host:bad/gh/o/r": parse "entire://alice:s3cret@host:bad/gh/o/r": invalid port ":bad" after host after:  fatal: invalid URL: invalid port ":bad" after host before: fatal: missing host in URL "entire://alice:s3cret@/x/o/r" after:  fatal: missing host in URL "entire://xxxxx:xxxxx@/x/o/r"

Unchanged on purpose: the push-too-large and low-disk messages still print exactly as before. The migrate plugin matches the push-too-large marker on git's stderr. Errors the helper does not recognise also still print in full, because then the chain is the only clue to the cause.

---

Note

Low Risk
Changes are mostly error presentation and error typing in the git remote helper; auth and transport behavior are preserved aside from clearer failover error precedence.

Overview
git-remote-entire now maps helper failures to short, actionable fatal: lines (cluster + repo, login hints, connectivity) via userMessage / statusMessage, while the full Go error chain is logged only under ENTIRE_DEBUG. Invalid or malformed URLs no longer echo userinfo; missing-host messages redact credentials in the displayed URL.

The remote transport exposes typed errors (HTTPStatusError, ErrNoNodeSucceeded, ErrUnreachable) so the helper can classify outcomes without parsing strings. Replica failover uses shared body parsing (HTML stripped from 5xx messages), and when cached replicas fail then the entry URL fails, the fallback status drives classification (e.g. entry 404 is not masked by a stale cached 503).

Minor UX: unrecognized GIT_PROTOCOL values are debug-logged instead of stderr warnings; Ctrl-C surfaces as interrupted. Push-too-large, low-disk, and request-policy errors stay verbatim for existing tooling.

^{Reviewed by Cursor Bugbot for commit a7f191d. Configure here.}

[pjbgf]

@cursor review

[Copilot]

🟡 Changes recommended

Error classification can hide final server responses and misreport security-policy rejections as connectivity failures.

2 open findings

• Prioritize fallback errors over stale cached replica failures · New
• Preserve unsafe-redirect diagnostics for exhausted replica requests · New

What changed in this PR

Makes git-remote-entire failures easier to understand while retaining detailed diagnostics in debug output.

Changes:

• Adds typed transport errors and concise user-facing messages.
• Redacts URL credentials and moves protocol warnings to debug output.
• Adds error-formatting and failover regression tests.

File	Description
internal/​remotehelper/​transport/​status_error_test.go	Tests typed failures and HTML suppression.
internal/​remotehelper/​transport/​proxy.go	Types exhausted-replica errors and filters response bodies.
internal/​remotehelper/​transport/​inforefs.go	Introduces typed HTTP status errors.
cmd/​git-remote-entire/​main.go	Formats failures, redacts credentials, and reduces stderr noise.
cmd/​git-remote-entire/​main_test.go	Tests messages, redaction, and protocol defaults.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[cursor]

Stale Bugbot comment from a previous run.

[pjbgf]

@cursor review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit a7f191d. Configure here.}