Machine learning technique to diagnose software crashes
The present disclosure describes a system and method for generating references for software crashes. The method includes applying a first machine learning model to a test case for a software application to adjust the test case to produce an adjusted test case and executing the adjusted test case on the software application to produce a software crash. The method also includes applying a second machine learning model to the software crash to generate a reference crash signature for the software crash and logging, using the second machine learning model, a reference crash configuration for the software crash.
1 . A method comprising:
applying a first machine learning model to a test case for a software application to adjust the test case to produce an adjusted test case;
executing the adjusted test case on the software application to produce a software crash;
applying a second machine learning model to the software crash to generate a reference crash signature for the software crash; and
logging, using the second machine learning model, a reference crash configuration for the software crash.
2 . The method of claim 1 , wherein adjusting the test case comprises adjusting at least one of a setting, an environmental variable, or data of the test case.
3 . The method of claim 1 , wherein adjusting the test case is based on a reward function that produces a reward when executing the adjusted test case on the software application produces the software crash.
4 . The method of claim 1 , wherein the reference crash signature comprises at least one of a call stack, a software variable, a memory dump, or a software setting of the software application during the software crash.
5 . The method of claim 1 , wherein the reference crash configuration comprises at least one of a global setting of a device that executed the software application during the software crash, an environmental variable of the device, or a model of the device.
6 . The method of claim 1 , further comprising storing, in a database, the reference crash signature and the reference crash configuration as a reference for the software crash.
7 . The method of claim 1 , wherein the first machine learning model implements at least one of a Bayesian model, natural language processing, image processing, a convolutional network, or a supervised learning model.
8 . The method of claim 1 , further comprising providing a plurality of reference crash signatures and a plurality of reference crash configurations to a third machine learning model to determine a proxy crash.
9 . A system comprising:
a memory; and
a processor communicatively coupled to the memory, wherein the processor:
applies a first machine learning model to a test case for a software application to adjust the test case to produce an adjusted test case;
executes the adjusted test case on the software application to produce a software crash;
applies a second machine learning model to the software crash to generate a reference crash signature for the software crash; and
logs, using the second machine learning model, a reference crash configuration for the software crash.
10 . The system of claim 9 , wherein adjusting the test case comprises adjusting at least one of a setting, an environmental variable, or data of the test case.
11 . The system of claim 9 , wherein adjusting the test case is based on a reward function that produces a reward when executing the adjusted test case on the software application produces the software crash.
12 . The system of claim 9 , wherein the reference crash signature comprises at least one of a call stack, a software variable, a memory dump, or a software setting of the software application during the software crash.
13 . The system of claim 9 , wherein the reference crash configuration comprises at least one of a global setting of a device that executed the software application during the software crash, an environmental variable of the device, or a model of the device.
14 . The system of claim 9 , wherein the processor stores, in a database, the reference crash signature and the reference crash configuration as a reference for the software crash.
15 . The system of claim 9 , wherein the first machine learning model implements at least one of a Bayesian model, natural language processing, image processing, a convolutional network, or a supervised learning model.
16 . A non-transitory computer readable medium storing instructions that, when executed by a processor, cause the processor to:
apply a machine learning model to a test case for a software application to adjust the test case to produce an adjusted test case;
executing the adjusted test case on the software application to produce a software crash;
generate a reference crash signature for the software crash; and
log a reference crash configuration for the software crash.
17 . The medium of claim 16 , wherein adjusting the test case comprises adjusting at least one of a setting, an environmental variable, or data of the test case.
18 . The medium of claim 16 , wherein adjusting the test case is based on a reward function that produces a reward when executing the adjusted test case on the software application produces the software crash.
19 . The medium of claim 16 , wherein the reference crash signature comprises at least one of a call stack, a software variable, a memory dump, or a software setting of the software application during the software crash.
20 . The medium of claim 16 , wherein the reference crash configuration comprises at least one of a global setting of a device that executed the software application during the software crash, an environmental variable of the device, or a model of the device.