13
1 Samsung Open Source Group Stefan Schmidt Samsung Open Source Group [email protected] EFL QA Statistics EFL Developer Day 2014

[E-Dev-Day 2014][10/16] QA Statistics

Embed Size (px)

Citation preview

Page 1: [E-Dev-Day 2014][10/16] QA Statistics

1Samsung Open Source Group

Stefan SchmidtSamsung Open Source Group

[email protected]

EFL QA Statistics

EFL Developer Day 2014

Page 2: [E-Dev-Day 2014][10/16] QA Statistics

2Samsung Open Source Group

Agenda

● Introduction

● Jenkins

● Coverity

● Clang Static Analyzer

● Unit Tests

● Phab Review

● Phab Bugs

● Miscellaneous

● Future

Page 3: [E-Dev-Day 2014][10/16] QA Statistics

3Samsung Open Source Group

Introduction

● Recording QA stats since 2013-07-15

● 15 months full of numbers might give

some insights

Page 4: [E-Dev-Day 2014][10/16] QA Statistics

4Samsung Open Source Group

Jenkins

● Almost all our QA tools are triggered from Jenkins

● Jobs are running on every repository change and

nightly

● Failure rates vary a lot

Page 5: [E-Dev-Day 2014][10/16] QA Statistics

5Samsung Open Source Group

Coverity

● Issue count got down lately due to Raster's Coverity addiction

● Elementary, Emotion generic players and Terminology already at 0

● Enlightenment, EFL and Evas generic loaders are not far away

● We should be able to bring them down to 0 for 1.12 or 1.13

● This will allow us to react only on newly created issues

Page 6: [E-Dev-Day 2014][10/16] QA Statistics

6Samsung Open Source Group

Clang Static Analyzer

● Nobody really takes a look at these

● Reported issues are harder to understand compared to Coverity

● No way to mark something as false positive in a web UI.

● I expected way more clang issues to vanish when Coverity issues got down

– Maybe that means clang reports still have some value for us

● Problems with understanding our hash_add, list_add function (leak reports)

Page 7: [E-Dev-Day 2014][10/16] QA Statistics

7Samsung Open Source Group

Unit Tests

● Less than 40% of the code base covered (with 400 tests)

● Biggest problem from a QA perspective

● What should we do to improve this situation?

– Man power to write tests covering the existing code

– Make test cases for new API's mandatory

– Usage of Exactness for UI testing

– Regression tests for fixed bugs?

Page 8: [E-Dev-Day 2014][10/16] QA Statistics

8Samsung Open Source Group

Phab Review

● Only 60 open from 1527 review requests

● Do not rely on the web UI only for review

– Always have a look at the output of git show before pushing

– Author not setup but phab ID shows instead

– White space changes not shown depending on UI setting (default should be fine now)

● Someone motivated to hook Phab up with jenkins?

– We could run new patches through jenkins and add feedback

Page 9: [E-Dev-Day 2014][10/16] QA Statistics

9Samsung Open Source Group

Phab Bugs

● 396 open bugs from 1722 reports

● 179 of them are whishlist items

Page 10: [E-Dev-Day 2014][10/16] QA Statistics

10Samsung Open Source Group

Miscellaneous

● Is the weekly report mail still interesting?

● Change something?

● More statistics?

– Would need a volunteer to write a script fetching

data from the web services

Page 11: [E-Dev-Day 2014][10/16] QA Statistics

11Samsung Open Source Group

Future

Potential improvements:

– Integrate ABI checker with jenkins to find problems earlier

– Try out more -fsanitize options from gcc and clang

– Volunteer for setting up qemu ARM builds?

– Volunteer for setting up expedite benchmarking on E3?

– Volunteer to finish exactness setup on jenkins?

– More testing on platforms not used by main devs (*BSD, Windows, Mac OSX)

QA work seems not to attract many people in the community.

Companies could help on this to reduce their internal QA

costs.

Page 12: [E-Dev-Day 2014][10/16] QA Statistics

Thank you.

12Samsung Open Source Group

Page 13: [E-Dev-Day 2014][10/16] QA Statistics

13Samsung Open Source Group

We are hiring.

Contact me if you want to hear details.

[email protected]