Re: Improve contributor experience and PR backlog

2016-03-06 Thread Rikki Cattermole via Digitalmars-d

A couple years ago I fixed a bug in dmd.
After like a month it turned out it actually fixed multiple bugs.

It was not merged. Nobody reviewed it past fixing the commits as one.
Many months later another PR was made to fix only one of the bugs and 
that was merged.


Funny how things go. I have not tried to fix any since.


Re: Improve contributor experience and PR backlog

2016-03-06 Thread Jack Stouffer via Digitalmars-d

On Sunday, 6 March 2016 at 16:44:31 UTC, Seb wrote:
Rather than the default mode for a change to be rejected until 
enough people sign off, we make the default for every change 
to land. This puts the onus on reviewers to note exactly what 
adjustments need to be made in order for it to land [4].


I really do like this idea. It could work if the review window 
is increased to maybe one week and everything without a "needs 
changes" tag (or "needs consensus" for major changes) is fine 
to merge then.


Unfortunately, this only works when the number of reviewers is 
high and they review everything. Phobos contributions have about 
four regular reviews and about ten people who review on and off. 
I can't blame people for not reviewing things at once, because 
the things that I review take a while for me to get through 
because I find the process a little tedious (not a D/Phobos 
specific thing, it's just about looking at other's code). I 
usually allow for two weeks before I ping someone because of this.


The PR of mine you referred to had serious problems on it's first 
implementation and should not have been pulled in 36 hours. There 
was even a whole form discussion about it: 
http://forum.dlang.org/post/n0p4s3$5em$1...@digitalmars.com


While I would love nothing more than for reviewing to go faster, 
I can't think of any mechanism to force reviews and contributors 
to respond faster to comments/make comments.


On a more positive note, trivial changes usually are merged 
within two days.