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
{{ message }}
Repository navigation
get_event_stacktrace returns nothing for exception-only events #2144
For an event with an exception entry but no threads entry (the common case for Python and most server SDKs), get_event_stacktrace returns only "No thread stacktraces were found for this event." The exception stacktrace that is in the same event is not returned.
fetchAndFormatEventStacktrace reads only the threads entry and returns early when it is empty (packages/mcp-core/src/tools/support/event-stacktrace.ts:46-55, getThreads at :73-77).
The tool description scopes it to threads (packages/mcp-core/src/tools/catalog/get-event-stacktrace.ts:23-39), but nothing tells the agent that exception frames are only in get_issue_details / get_sentry_resource, so agents call it for exception events and get an empty answer.
Reported through MCP user feedback (MCP-SERVER-GN2) on a Python exception event. The call itself was not in sampled traces; the behavior is from the code path above.
I went through the reported behavior and plan to update the stacktrace extraction logic to handle exception-only events when the threads entry is absent or empty.
I'll also add regression tests covering exception-only events, thread-based stacktraces, and events without any stacktrace data.
Please let me know if there are any implementation considerations or preferred approaches before I start.
Another report hit this: MCP user feedback MCP-SERVER-GRJ (Android SIGABRT, native event). get_event_stacktrace returned no_threads (packages/mcp-core/src/tools/support/event-stacktrace.ts:99), so the agent couldn't get the crashing native stack from it. The broader native-diagnostics gap (debug images, frame addresses, native processing errors, signal) is tracked in #2357.
Related, separate path: MCP user feedback MCP-SERVER-GS8 hit get_event_stacktrace returning no_threads on native message events whose stack is in the event-level stacktrace entry, not exceptions or threads. That's tracked in #2401, and the draft fix in #2402 touches the same no_threads branch in packages/mcp-core/src/tools/support/event-stacktrace.ts. It doesn't return exception stacks, so this issue stays open. Anyone fixing this one should rebase on that PR to avoid a conflict.
Activity
linear-code commentedon Nov 10, 2026
MCP-1207
launchforgedev commentedon Nov 13, 2026
Hi! I'd like to work on this issue.
I went through the reported behavior and plan to update the stacktrace extraction logic to handle exception-only events when the
threadsentry is absent or empty.I'll also add regression tests covering exception-only events, thread-based stacktraces, and events without any stacktrace data.
Please let me know if there are any implementation considerations or preferred approaches before I start.
sentry-junior commentedon Nov 18, 2026
Another report hit this: MCP user feedback MCP-SERVER-GRJ (Android SIGABRT, native event).
get_event_stacktracereturnedno_threads(packages/mcp-core/src/tools/support/event-stacktrace.ts:99), so the agent couldn't get the crashing native stack from it. The broader native-diagnostics gap (debug images, frame addresses, native processing errors, signal) is tracked in #2357.sentry-junior commentedon Nov 18, 2026
Related, separate path: MCP user feedback MCP-SERVER-GS8 hit
get_event_stacktracereturningno_threadson native message events whose stack is in the event-levelstacktraceentry, not exceptions or threads. That's tracked in #2401, and the draft fix in #2402 touches the sameno_threadsbranch inpackages/mcp-core/src/tools/support/event-stacktrace.ts. It doesn't return exception stacks, so this issue stays open. Anyone fixing this one should rebase on that PR to avoid a conflict.