A research study on abandoned code reviews

[Jan Iversen]

Hi

Copying our mailing list, to make the answer publicly available.

Sounds like an interesting topic, but looking at the one you asked me about it seems to me, you look at the abandoned ones only in a statistical sense, that will not do. I am actually curious, why do you look at abandoned patches, the right question (at least to me) is “why do some patches not get merged”, for LibreOffice the result is the same, but for other projects it is very different.

33697 was submitted and abandoned by myself, because it had some problems…you will see a lot of people submit patches, to get them tested on our build boxes, and if there is a problem they either submit a second patch set or (as in my case) abandon the patch and submit a new one.

I trust you are also aware that TDF (in contrary to most other foundations) actively monitor gerrit. Patches who are untouched for 4 weeks, get a “polite ping”, and 4 weeks later still untouched gets abandoned. This procedure gives a lot more abandoned patches compared to other projects.

If you speak generally you need to look more carefully at the reasons.

1) Submitter abandons patch. This is quite natural as described above.

2) People with committer status, typically submit a patch and forget it, so the system catches it and they typically abandon it, when they see the “polite ping”.

3) New people submit a patch (some without having compiled the code) receive a review and decide not to continue. This is only true for first time patches.

4) New people do a good job of making their patch work, but get review comments e.g. due to the user experience not being correct, and stop working

Again compared to other projects, we ensure that all patches get reviewed, so no patch is abandoned due to lack of review (something I see happening in another Foundation I work with).

Have fun with your study, and please share the outcome on our dev list, and I think it will be interesting reading.

rgds
jan I.


> On 10 Mar 2017, at 07:00, wqyy at zju.edu.cn wrote:
> 
> Dear jan, 
> 
> Sorry to bother you!
> 
> I am a Ph.D student in Zhejiang University, China. Currently, we are doing a research study on identifying root causes of abandoned code reviews. As we noticed, code reviews are abandoned for various reasons, e.g., build failed, wrong branch, duplicate patch, and so on. 
> 
> Since you are an expert in libreoffice, and I'm writing to ask you the reason why the following code review was abandoned (https://gerrit.libreoffice.org/#/c/33697/ <https://gerrit.libreoffice.org/#/c/33697/>). Also, it would be great if you can tell us some more reasons why a code review is abandoned:-) I would be really grateful if you could reply me.
> 
> Please kindly advice.
> 
> Thank you very much for your help and happy holidays!
> 
> Yours sincerely,
> 
> Qingye Wang

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice/attachments/20170310/d7f4412a/attachment.html>