Module transport 

HTTP transport shared by every forge adapter.

Retrying, timing out, paging, and mapping a status to a failure class are the same problem for Forgejo and GitHub; only the paths and payloads differ. Adapter-specific status meaning (for example, that Forgejo answers 403 on branch protection create when the rule already exists) is resolved by the one call site that needs it, by matching on the TransportError variant that call produced.

Structs

Http

Shared HTTP transport for a forge’s REST API.

RequestLabel

The method and path of a request, as it appears in an error.

RetryPolicy

How a retryable failure is retried before it is reported.

TransportBuildError

Failure building the underlying HTTP client.

Enums

TransportError

Anything that stops a request from producing a usable answer.

Constants

DEFAULT_PAGE_LIMIT 🔒

The most pages a listing may span unless Http::with_page_limit says otherwise.

MAX_RATE_LIMIT_WAIT 🔒

The longest wait a rate-limited request honours. A forge that asks for more makes the request fail with TransportError::RateLimitWaitTooLong.

RATE_LIMIT_REMAINING 🔒

RATE_LIMIT_RESET 🔒

Functions

log_retry 🔒

Log a retryable failure before the attempt that follows it.

rate_limit_wait 🔒

How long the response asks the caller to wait before retrying.

retry_wait 🔒

The wait before retrying error, which is the wait the forge stated or else backoff. A stated wait above the cap fails the request.

retryable 🔒

A rate limit is retried for every method, because the forge refused the request without applying it. Anything else is retried only for an idempotent method, because it does not tell whether the forge applied it.

summarize

Trim a forge error body so a huge HTML page cannot flood the output.