Citation Context

...em. Even if these events get processed sequentially in one thread, most events are logically concurrent to each other, as they may not be ordered by any programmer specified ordering operations. Unfortunately, most existing tools such as data-race detectors [11] for finding concurrency errors assume a thread-based model. Naively applying these tools for event-driven mobile systems works poorly, because they implicitly assume that events handled in one thread are ordered by the program order. Recent work proposed a causality model for detecting races in the event-driven JavaScript applications [18, 20]. However, a causality model that accounts for the unique causal relations in a mobile system, and a race detector based on it has been lacking. This paper presents the first race detection tool named CAFA for mobile applications. We chose to implement CAFA for the widely used Android platform [1], as the system and many of its applications are open-source. An Android application uses a special thread, named the looper thread, to select events from an event queue one at a time, and processes them sequentially by executing their handlers. There may be additional regular threads that communicate...

Citation Context

...e can be integrated with Asynchronizer to help developers avoid GUI access in AsyncTask while performing refactoring. Some techniques are proposed to analyze event-driven applications. Raychev et al. =-=[31]-=- proposed a dynamic data race detection for event-driven program based on Javascript. Zheng et al. [39] statically check concurrency bugs due to asynchronous calls for Ajax applications. Less scalable...

Citation Context

...ching algorithm to avoid such mistakes. 5. RELATEDWORK Several dynamic analyses for JavaScript have been proposed recently, e.g., TypeDevil [42] detects type inconsistencies in JavaScript, EventRacer =-=[44]-=- identifies event-related data races, and Mesbah et al. check for cross-browser issues [39]. JITProf is a profiler that identifies performance bottlenecks caused by code that JIT engines cannot effect...

Citation Context

...king conventional threads, the presence of asynchrony creates a potential for races, in a way that is similar to concurrent code running on a single processor. In particular, the ordering of event execution in JavaScript as well as the timing of completions of asynchronous requests is non-deterministic, prone to be affected by network timing, etc. Asynchrony in JavaScript programs, therefore, may result in concurrency-related errors. Prior work: Zheng et al. [9] propose a static technique to detect potential races in JavaScript applications. More recently, Petrov et al. [6] and Raychev et al. [7] have observed the potential for asynchrony creating out-of-order execution and developed a notion of race conditions for Web applications written in JavaScript. In principle, race conditions can arise because of accesses to data shared among components of a Web page which are not ordered by proper synchronization, or, more formally, a happens-before relation. Of course, on a Web page, the entire DOM is (giant blob of) global state, creating the potential for races. Petrov et al. [6] define a happens-before relation for Web pages and generalize the notion of race conditions to take into accoun...

Citation Context

... be executed in one order, but due to a programming mistake are executed in the reverse order. The order violation coincides with the race bug pattern defined in Petrov et al. [17] and Raychev et al. =-=[19]-=-. Figure 2 shows an order violation example. main.html dynamically loads the external file (extn.js) (lines 3–5) that contains fn() function definition (line 11). Thus, invoking fn() should not occur ...

Citation Context

...d to the controller at any time, all non-deterministically. Highly asynchronous programs are prone to concurrency errors caused by interfering access (a form of data races) to shared memory locations =-=[1]-=-, and SDN control software is no exception. However, it is well known that discovering those errors is difficult as they often depend on a particular ordering of specific events. At the same time, det...

Citation Context

...aychev et al. observe that race detection techniques such as the one of Petrov et al. report an overwhelming number of races in cases where event handler execution is coordinated via shared variables =-=[29]-=-. They reduce the number of false positives by introducing a notion of race coverage where a race a covers a race b if treating a as synchronization eliminates b as a race. Raychev et al. implemented ...

Citation Context

...ers have recognized the need for automatically discovering EAs. For instance, CAFA [29] identifies Use-After-Free (UF), a common type of EA in Android applications, while WEBRACER [47] and EVENTRACER =-=[50]-=- focus on detecting certain types of EAs in client-side web applications. However, these techniques are based on dynamic analysis; therefore, they offer no guarantees Permission to make digital or har...

Citation Context

... Triage mode represents a middle ground between these two extremes, where Guardrail performs a best-effort estimation of the safety of completing the I/O operation by automatically triaging the fault =-=[22, 27, 34]-=-. If the I/O operation is deemed safe, Guardrail behaves like permissive mode, otherwise it behaves like stringent mode. Although this flexibility allows Guardrail to be configured in interesting ways...