flacocobot support for fault-localization in Repairnator

[andre15silva]

Adds support for running flacocobot on PRs and as a runnable

[andre15silva]

Next steps:

• add support for scanning pull-requests (GithubPullRequestScanner) #1244 (@dginelli )
• Adapt FlacocoScanner to use the new GithubPullRequestScanner (@dginelli )
• Adapt DockerPipelineRunner to the new GithubInputBuild format for PRs (@andre15silva )
• Adapt the FAULT_LOCALIZATION pipeline mode to receive a PR and release the docker image version (@andre15silva )

[andre15silva]

We should update github-api: https://mvnrepository.com/artifact/org.kohsuke/github-api

[andre15silva]

@dginelli I don't remember if I've told you this, but please keep in mind that the personal access token you use must have the public_repo scope to be able to create pull request reviews.

It's really easy to forget that it is the case, and the error message is really unclear.

See: https://github.heygears.community/t/404-while-creating-new-pr-review/196660

[andre15silva]

I've pushed a new version of the repairnator/pipeline image, 3.4.

It now supports the fault localization mode: docker run -e REPAIR_MODE=FAULT_LOCALIZATION -e GITHUB_URL=https://github.com/repairnator/failingProject -e GITHUB_PR=7 -e FLACOCO_THRESHOLD=0.12 -e GITHUB_OAUTH=$GITHUB_OAUTH repairnator/pipeline:3.4

WIP on Adapt DockerPipelineRunner to the new GithubInputBuild format for PRs

[dginelli]

@dginelli I don't remember if I've told you this, but please keep in mind that the personal access token you use must have the public_repo scope to be able to create pull request reviews.

It's really easy to forget that it is the case, and the error message is really unclear.

See: https://github.heygears.community/t/404-while-creating-new-pr-review/196660

Thank you for the information!

[andre15silva]

@dginelli You should now be able to run flacocobot.

This comment, repairnator/failingProject#7 (review), was generated through the FlacocoScanner.

Let me know if you need help or have any doubt.

[dginelli]

@dginelli You should now be able to run flacocobot.

This comment, repairnator/failingProject#7 (review), was generated through the FlacocoScanner.

Let me know if you need help or have any doubt.

Hi @andre15silva,

Thank you for the work, I managed to run flacocobot and to push its suggestions on a pull request, but I have a couple of doubts about its functioning.

The first one is related to the volumes: I had to change this part of the source code adding these lines

final Volume repairnatorWorkspace = Volume.builder()
        .name("repairnator_workspace")
        .driver("local")
        .build();
final Volume repairnatorWorkspaceCreated = docker.createVolume(repairnatorWorkspace);

final Volume repairnatorLogs = Volume.builder()
        .name("repairnator_logs")
        .driver("local")
        .build();
final Volume repairnatorLogsCreated = docker.createVolume(repairnatorLogs);

before the ones that I linked above in order to create the volumes, otherwise the pipeline was not able to run since it was not able to find the volumes. I don’t know if it is a problem that occurs only to me or not (I tried it on my Mac and on a clean virtual machine with Ubuntu 18.04). Before running flacocobot, I downloaded the repairnator/pipeline image.

The second aspect is related to the projects that have more than one failed open pull requests. When such a situation occurs, I noticed that the pipeline crashes. I think this is related to the fact that the pipeline tries to work on the same folder (since the repository is the same), and it fails to clone the repository.

Attached to this comment, you can find an example of the stack trace that I get when two pull requests for the same repository are available.

Also in this case, I don’t know if you already tried this test case without having errors or I’m doing anything wrong. If you have the error too, maybe a possible solution would to clone a repository in a folder that has the name of the repository followed by the id of the commit associated with the pull request, so that the folders have always unique names.

exception.txt

Thank you!

[andre15silva]

Hi @andre15silva,

Hi @dginelli !

The first one is related to the volumes: I had to change this part of the source code adding these lines

final Volume repairnatorWorkspace = Volume.builder()
        .name("repairnator_workspace")
        .driver("local")
        .build();
final Volume repairnatorWorkspaceCreated = docker.createVolume(repairnatorWorkspace);

final Volume repairnatorLogs = Volume.builder()
        .name("repairnator_logs")
        .driver("local")
        .build();
final Volume repairnatorLogsCreated = docker.createVolume(repairnatorLogs);

before the ones that I linked above in order to create the volumes, otherwise the pipeline was not able to run since it was not able to find the volumes. I don’t know if it is a problem that occurs only to me or not (I tried it on my Mac and on a clean virtual machine with Ubuntu 18.04).

Oh, I created the volumes manually when I ran it. I forgot to tell you, sorry.

The second aspect is related to the projects that have more than one failed open pull requests. When such a situation occurs, I noticed that the pipeline crashes. I think this is related to the fact that the pipeline tries to work on the same folder (since the repository is the same), and it fails to clone the repository.

Attached to this comment, you can find an example of the stack trace that I get when two pull requests for the same repository are available.

Also in this case, I don’t know if you already tried this test case without having errors or I’m doing anything wrong. If you have the error too, maybe a possible solution would to clone a repository in a folder that has the name of the repository followed by the id of the commit associated with the pull request, so that the folders have always unique names.

I didn't try it with project with more than 1 failing PR.

I'll fix that in a moment, to have a unique directory name everytime we clone.

[dginelli]

Oh, I created the volumes manually when I ran it. I forgot to tell you, sorry.

No problem, don't worry! ;-) I didn't push the changes to create the volumes on the repo, but I can add them if you think that it is a good idea.

I didn't try it with project with more than 1 failing PR.

I'll fix that in a moment, to have a unique directory name everytime we clone.

Ok, great! :-) Meanwhile, I continue to test the scanner with other possible cases that come to my mind.

[andre15silva]

@dginelli Conflict fixed!

I've updated the docker image too. The code should pull the new image automatically, otherwise just pull it again: docker pull repairnator/pipeline:3.4

Anything else let me know :)

[dginelli]

Hi @andre15silva,

I finished the implementation part and I updated the documentation related to the use of Flacocobot. I also added a couple of test cases to test the GithubPullRequestScanner, that is the scanner used by Flacocobot to detect the pull requests associated with the projects under analysis.

When you have time, you can check if everything is ok :-)

[monperrus]

Great, thanks Davide! @andre15silva <https://github.com/andre15silva> WDYT?

[andre15silva]

Great, thanks Davide! @andre15silva <@andre15silva> WDYT?

Oh my, this thread got buried sorry @dginelli !

I'm busy until Friday, but as soon as I'm done I'll go over the changes and finish what's left (tests and docker at least).

[andre15silva]

• tests to be updated

[andre15silva]

Ready for review @monperrus , the remaining tasks need to be done after we merge this (#1248)

[monperrus]

Great, thanks a lot!