Watcher: precise and fully-automatic on-site failure diagnosis

A system comprising:

• a processor, memory accessible by the processor and adapted to store program instructions, and the program instructions stored in the memory and executable by the processor to perform:
• recording information relating to execution of software code;
• intercepting a failure signal indicating a failure detected in the execution of the software code;
• identifying a failing register and an instruction pointer associated with the failure signal to determine a location of the failure in the software code;
• identifying all possible branches of execution at the location of the failure in the software code and placing breakpoints on corresponding jump instructions to determine a failing branch;
• replaying the execution of the software code based on the recorded information;
• determining the failing register using a failing instruction pointed to by the instruction pointer;
• determining, using the breakpoints, a last assignment to the failing register before the failure;
• if at the breakpoints during the replay of the execution of the software code a value of the last assignment associated with the failing register is determined to come from an additional register, setting an additional breakpoint and replaying the execution of the software code based on the recorded information and the additional breakpoint;
• and
• if at the breakpoints during the replay of the execution of the software code the value of the last assignment associated with the failing register is determined not to come from the additional register, generating a failure report indicating the failing register and the failing instruction as a root cause of the failure detected in the execution of the software code.

A computer program product comprising a non-transitory computer readable storage having program instructions embodied therewith, the program instructions executable by a computer system, to cause the computer system to perform a method comprising:

• recording, at the computer system, information relating to execution of software code;
• intercepting, at the computer system, a failure signal indicating a failure detected in the execution of the software code;
• identifying, at the computer system, a failing register and an instruction pointer associated with the failure signal to determine a location of the failure in the software code;
• identifying, at the computer system, all possible branches of execution at the location of the failure in the software code and placing breakpoints on corresponding jump instructions to determine a failing branch;
• replaying the execution of the software code based on the recorded information;
• determining, at the computer system, the failing register using a failing instruction pointed to by the instruction pointer;
• determining, at the computer system, using the breakpoints, a last assignment to the failing register before the failure;
• if at the breakpoints during the replay of the execution of the software code a value of the last assignment associated with the failing register is determined to come from an additional register, setting an additional breakpoint and replaying the execution of the software code, based on the recorded information and the additional breakpoint;
• and
• if at the breakpoints during the replay of the execution of the software code the value of the last assignment associated with the failing register is determined not to come from the additional register, generating, at the computer system, a failure report indicating the failing register and the failing instruction as a root cause of the failure detected in the execution of the software code.