Avaliar o efeito dos handlers

[pauloborba]

document analyticall detailing handlers effects: https://docs.google.com/document/d/1cMwIxbeW8SwZzMWvMdkWEsShOdIRPuahBDtOcmgyqxc/edit?usp=sharing

we have now to assess that empirically

[pauloborba]

main points for the study:

1. deal with lack of groundtruth. possible actions: use merge result in the repository as an approximation (string comparison), assessment/pattern scripts, experts opinion/small sample. check TGs (Giovanni and Alice)

2. use mining framework

3. invoke tool with different parameters (handlers) and compare results

[pauloborba]

check #23

[pauloborba]

to deal with ground truth, use travis info (as in the ASE paper), and leuson's tool for build conflicts.

[jvcoutinho]

About this, to confirm: if both handlers output a conflict -> nothing to be done if one handler outputs a conflict and the other doesn't -> check if there's a build crash, if false, manual check if both handlers don't output a conflict -> nothing to be done Am I missing something (or there is something wrong), @pauloborba?

[pauloborba]

just a correction @jvcoutinho: if both handlers output the same conflict -> nothing to be done

[pauloborba]

@jvcoutinho refine and document algorithms, as discussed today

and then increase the sample so that we can have greater confidence in the results

if you have developed new services for the mining framework, please send a PR