Citation Context

...enable more freedom of implementation, which is particularly important for interrupts. The next section covers the use of monitors in Mesa. There are also tools for dynamic detection of races. Eraser =-=[44]-=- detects unprotected shared variables using a modified binary. “On the fly” race detectors [31, 39] serialize all accesses to a variable to verify serializability. These approaches only catch errors t...

Citation Context

...but will maintain some selected information about the past, and either do analysis of this information on-the-fly, or after program termination. An example is the data race detection algorithm Eraser =-=[38]-=- developed at Compaq. Another example is a locking order analysis called LockTree which we have developed. Both these algorithms have been implemented in JPF. Below we describe these two algorithms, a...

Citation Context

...evel compilation. Many projects have hard wired application-level information in compilers. These projects include: compiler-directed management of I/O [24]; the ERASER dynamic race detection checker =-=[30]-=-; ParaSoft's Insure++ [19], which can check for Unix system call errors; the use of static analysis to check for security errors in privileged programs [1]; and the GNU compilers' -Wall option, which ...

Citation Context

...ur rule templates, then we use statistical analysis to rank the derived rules. High-level Compilation. Many projects have embedded hard-wired application-level information in compilers to find errors =-=[1, 3, 4, 8, 21, 24]-=-. These projects find a fixed set of errors, whereas we derive new checks from the source itself, allowing detection of a broader range of errors. The checking information we extract could serve as in...

Citation Context

...practices like uninitialized memory reads, memory leaks and array bounds overruns. However, most of these errors are automatically avoided in a type−safe, garbage−collected language like Java. Eraser =-=[18]-=- is a dynamic analysis system which detects the set of locks protecting each variable in a multithreaded program and flags inconsistencies in the usage of locks. Various dynamic techniques have been p...

Citation Context

...roving programming productivity is served also by better tools supporting traditional testing and runtime checking. It is plausible that a systematic, disciplined use of a dynamic checker like Eraser =-=[43]-=- would do as well as ESC at detecting race conditions and deadlocks. Also in the area of detecting race conditions is the Cilk tool Nondeterminator-2 [2], which is based on an intriguing combination o...

Citation Context

...ssable location; ffl Cachegrind: A cache profiler that supports line-by-line source annotations of instruction and data cache misses; ffl Helgrind: A data-race detector that uses the Eraser algorithm =-=[25]-=-; ffl Nulgrind: the "null" skin that performs no instrumentation. 1.3 Contributions of This Work We claim it is easier to write powerful supervision tools with Valgrind than with other, similar progra...