Machine learning technique to diagnose software crashes

1. A method comprising:

receiving a crash signature and a crash configuration, the crash signature generated in response to a software crash in a software application that occurred using the crash configuration; and

applying a first machine learning model to determine a reference of a plurality of references that is closest to the crash signature and the crash configuration, the reference comprising a reference crash signature and a reference crash configuration, the reference generated based on a proxy crash, wherein the proxy crash was generated prior to the software crash by executing a modified test case against the software application.

2. The method of claim 1 , further comprising ranking the plurality of references based on their distances to the crash signature and the crash configuration after applying the first machine learning model.

3. The method of claim 1 , further comprising applying a second machine learning model to the crash configuration to determine a workaround for the software crash.

4. The method of claim 3 , wherein the second machine learning model uses a reward function to determine the workaround.

5. The method of claim 1 , further comprising adjusting, by another machine learning model, a configuration of a test case to produce the modified test case.

6. The method of claim 5 , further comprising, executing, by the another machine learning model, the modified test case against the software application to generate the proxy crash.

7. The method of claim 1 , further comprising, generating, by an additional machine learning model, the reference crash signature and the reference crash configuration based on the proxy crash.

8. The method of claim 1 , wherein the crash signature comprises a crash stack or a memory dump of the software crash.

9. An apparatus comprising:

a memory; and

a hardware processor communicatively coupled to the memory, the hardware processor configured to:

receive a crash signature and a crash configuration, the crash signature generated in response to a software crash in a software application that occurred using the crash configuration; and

apply a first machine learning model to determine a reference of a plurality of references that is closest to the crash signature and the crash configuration, the reference comprising a reference crash signature and a reference crash configuration, the reference generated based on a proxy crash, wherein the proxy crash was generated prior to the software crash by executing a modified test case against the software application.

10. The apparatus of claim 9 , the hardware processor further configured to rank the plurality of references based on their distances to the crash signature and the crash configuration after applying the first machine learning model.

11. The apparatus of claim 9 , the hardware processor further configured to apply a second machine learning model to the crash configuration to determine a workaround for the software crash.

12. The apparatus of claim 11 , wherein the second machine learning model uses a reward function to determine the workaround.

13. The apparatus of claim 9 , the hardware processor further configured to adjust, by another machine learning model, a configuration of a test case to produce the modified test case.

14. The apparatus of claim 13 , the hardware processor further configured to execute, by the another machine learning model, the modified test case against the software application to generate the proxy crash.

15. The apparatus of claim 9 , the hardware processor further configured to generate, by an additional machine learning model, the reference crash signature and the reference crash configuration based on the proxy crash.

16. The apparatus of claim 9 , wherein the crash signature comprises a crash stack or a memory dump of the software crash.