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
* initial queue attempt
* add timer
* prototype double buffer approach
* update logs unit tests for batching
* replace timer with bgworker
* add first integration tests
* update example.c with correct log thread amounts
* cleanup
* add wait for 'adding' logs in flush logic
* go back to single queue for performance testing
* add time checks
* add ToDos + cleanup sentry_value_t cloning
* initial attempt
* cond_wait for timer + 'adding' spinlock
* add sleep for tests
* add sleep for tests
* force flush before attempting timer_worker shutdown
* add proper cond_wait for 'adding' counter
* revert to manual flush on shutdown instead of timer thread flush
* add separate timer_stop atomic
* cleanup + replace 'adding' cond_wait by pure spinlock
* change bgworker for simpler thread implementation
* cleanup
* fix memleak
* fixes
* cleanup
* cleanup
* windows fixes
* update shutdown order
* fixes
* explicit check to stop timer task
* windows cleanup
* loosen threaded test assertion for CI
- too much variability in thread scheduling, so we can expect pretty much anything
* add continue for unexpected logs flush trigger instead of attempting flush
* Windows re-add condition variable trigger case
* feat(logs): add `before_send_log` (#1355)
* add `before_send_log` callback
* add `before_send_log` callback tests
* (temporary) add debug for calling sentry_options_free
* remove early return
* add late return
* cleanup
* fix ownership issues in single buffer batching (#1362)
* let the producer thread sleep for 1ms between logs
* fix two missing NULL checks in the json serializer
* clean up logging and early exits in `enqueue_log_single()`
* clean up ownerships in logs
* eliminate clones (we expect that everything outlives the logs being sent except local construction)
* use incref everywhere where we ref global state. this was the cause of the UAF, partially solved with the clones but a few were missing. no reason to clone if we do not want to disconnect for a particular object graph
* clarify that add_attribute expects ownership
* minimize scope_mut by moving os_context out
* raise that log output in throughput tests add to variability (stdout logging should be turned off when running a limit)
* log error in case we weren't able to start the log batching worker
* fix clang-cl warning
* ci: fix failing mingw build (#1361)
* ci: fix failing mingw build
* split `ASM_MASM_COMPILER` and `_FLAGS`
* add `ASM_MASM_FLAGS` in `mingw` install step
* specify the `CMAKE_ASM_MASM_COMPILER` as a `FILEPATH`
* clean up CMAKE_DEFINES construction so it is easier to diff in the future
* fix `LLVM_MINGW_INSTALL_PATH` to be referenced locally rather than $env
(cherry picked from commit 519554f)
* use UNREACHABLE macro to fix anal warnings
* batching double buffered (#1365)
* first attempt at double buffered
* remove the sleep from the windows thread func
* clean up thread waiting in the example
* adapt the double buffer to use retries, fix remaining issues, clean up and write inline docs
* return early in example on sentry_init error.
* fix formatting via shorter name for thread gate atomic
* improve inline docs of log_buffer_t members
* fetch os_context from scope
* move scope/options data retrieval into separate function + add expected keys to test
* update logs API to return status code
* cleanup
* add log-event trace connection test
* remove duplicate test
* specify macOS SDKROOT
---------
Co-authored-by: JoshuaMoelans <[email protected]>
* add flush retry for missed flush requests
* move flush retry into flush function
* add docs
---------
Co-authored-by: Mischan Toosarani-Hausberger <[email protected]>
0 commit comments