fix(logging): prevent log errors from crashing main thread

Fixes #2793

When log events are shipped to Sentry (either synchronously when the
buffer is full, or via the background thread), network errors could
propagate to the main application thread and crash the application.

Author: solnic

sentry-ruby/lib/sentry/log_event_buffer.rb

@@ -69,6 +69,10 @@ def size
 
     def send_events
       @client.send_logs(@pending_events)
+      @pending_events.clear
+    rescue => e
+      log_debug("[LogEventBuffer] Failed to send logs: #{e.message}")
+
       @pending_events.clear
     end
   end

sentry-ruby/spec/sentry/log_event_buffer_spec.rb

@@ -74,4 +74,61 @@
       expect(log_event_buffer).to be_empty
     end
   end
+
+  describe "error handling" do
+    let(:max_log_events) { 3 }
+
+    let(:error) { Errno::ECONNREFUSED.new("Connection refused") }
+
+    context "when send_logs raises an exception" do
+      before do
+        allow(client).to receive(:send_logs).and_raise(error)
+      end
+
+      it "does not propagate exception from add_event when buffer is full" do
+        expect {
+          3.times { log_event_buffer.add_event(log_event) }
+        }.not_to raise_error
+      end
+
+      it "does not propagate exception from flush" do
+        2.times { log_event_buffer.add_event(log_event) }
+
+        expect {
+          log_event_buffer.flush
+        }.not_to raise_error
+      end
+
+      it "logs the error to sdk_logger" do
+        3.times { log_event_buffer.add_event(log_event) }
+
+        expect(string_io.string).to include("Failed to send logs")
+      end
+
+      it "clears the buffer after a failed send to avoid memory buildup" do
+        3.times { log_event_buffer.add_event(log_event) }
+
+        expect(log_event_buffer).to be_empty
+      end
+    end
+
+    context "when background thread encounters an error" do
+      let(:max_log_events) { 100 }
+
+      before do
+        allow(client).to receive(:send_logs).and_raise(error)
+      end
+
+      it "keeps the background thread alive after an error" do
+        log_event_buffer.add_event(log_event)
+        log_event_buffer.start
+
+        thread = log_event_buffer.instance_variable_get(:@thread)
+
+        expect(thread).to be_alive
+        expect { log_event_buffer.flush }.not_to raise_error
+        expect(thread).to be_alive
+      end
+    end
+  end
 end