End-User Tracing for Software Replay

目前，大多数计算机故障都是由软件缺陷引起的。不幸的是，当前的软件开发文化容忍了存在大量缺陷的软件。 解决这一现象的最好办法是进行更彻底的测试。 该项目旨在增加现代软件所经历的测试和有效调试的数量。 这项研究使最终用户能够通过自动跟踪程序的执行并将跟踪作为错误报告提交来充当有效的测试人员。 这样的跟踪包含了软件开发人员重现故障所需的所有信息，即使该故障是不确定的。 这个跟踪和重放系统的目标反映了它所针对的最终用户环境。 跟踪系统应该有足够低的开销，以便在正常使用期间保留。 跟踪系统应该对跟踪进行翻译或编码，以维护用户的隐私。跟踪和重放系统应该易于最终用户使用;例如，它不需要修改内核或系统库。 如果成功，最终的系统将大大增加有效测试人员的基础，并有望产生更可靠的软件。

Software bugs currently cause the large majority of computer failures.Unfortunately, the current culture of software development tolerates therelease of software with numerous bugs. The best solution to this phenomenonis more thorough testing. This project seeks to increase the amount of testingand effective debugging undergone by modern software. This research enablesend users to serve as effective testers by tracing a program's executionautomatically and submitting the trace as a bug report. Such traces containall information needed for the software developer to reproduce the failure,even if that failure is non-deterministic. The goals of this trace-and-replaysystem reflect the end-user environment it is targeted for. The tracing systemshould have low-enough overhead to be left on during normal use. The tracingsystem should translate or encode the trace to maintain the user's privacy.The trace-and-replay system should be easy to use by the end user; for example,it should not need to modify the kernel or system libraries. If successful,the resulting system will dramatically increase the base of effective testersand hopefully result in more reliable software.