new socket-wrapper test case and questions around how we ensure quality with the socket wrapper code

[Andreas Schneider]

On Wednesday, May 11, 2016 08:32:10 AM Richard Sharpe wrote:
> Hi folks,
> 
> Attached is my now working test case for what net ads dns
> gethostbyname is doing.
> 
> Along the way, I realized that the testing infrastructure is very
> important and it is important to ensure that the test cases are
> correct.
> 
> So, I tested:
> 
> 1. Outside the socket wrapper tests by running my test stand-alone,
> and it succeeds.
> 
> 2. Under the current socket wrapper code without my patch, and it
> fails, as I expect.
> 
> However, if I send in a single patch that contains both the new test
> case and the fix, whoever merges the code cannot verify that the test
> case is worthwhile.
> 
> On the other hand, if I send in two patches, and someone merges them
> both and pushes, there will be two commits between which the tests
> will fail.
> 
> Perhaps I should submit three patch files. One with my new test case,
> one with the fixes for the socket wrapper code, and one squashed.
> 
> Any comments?

Please send two patches, the first fixing the issue and the second providing 
the testcase.

I normally apply them run the tests, call 'git revert' on the fix command and 
test it, then if it fails, remove the change and commit it. This way git 
bisect keeps working :)


Cheers,


	-- andreas