You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The SHA-256 helper no longer loads an unversioned libcrypto on macOS. httputil.rkt's native SHA-256 shortcut opened bare libcrypto; current
macOS releases make dyld abort the whole process on the unversioned /usr/lib/libcrypto.dylib stub, which killed every process that merely
required the module (the test suite included) — the exception handler
never sees a dyld-level abort. macOS now resolves versioned names only
(Racket's bundled libcrypto.3); Linux keeps the unversioned fallback.
The first-run rslsync installer works again: http-fetch ran bytes->string/utf-8 over the status value that get-pure-port/headers#:status? #t already returns as a string, so every successful 200 response died on a contract violation — the download card failed unconditionally on every machine, with the retry loop dressing it up as a network error ("check your network connection"). The status parsing now accepts the string shape (bytes kept as a guard for older net/url revisions) in a response-code helper covered by its own tests.
The embedded official UI renders again in desktop mode: the desktop bridge's callapi method answered a hardcoded null ("no entries are reachable in this UI build") — true for the UI it was written against, false since the daemon's Web UI moved its whole data plane into the desktop bridge (with window.btsync present, the UI's jQuery.ajax override routes every /gui/ and /api/ request through callapi and never speaks HTTP to the proxy, so the stub left the interface with no data at all: a blank window). callapi now forwards {action, params, type} to the same origin the web view loads — the auth-injecting proxy — over a libsoup session and answers with the daemon's response; transport and HTTP failures surface as {"error": …} results rather than rejected promises (a rejection crashes the UI's request wrapper on a missing .error field). The request finishes when the daemon answers, so the ~30s core-events long poll passes through untouched.