Re: [PATCH 0/6] Tracking setup improvements: per-remote config, autosetuppush

I haven't read any of these patches, other than applying them on top of master and looking at the resulting diff in the t/ directory in the aggregated form, and also noticing some style deviations in the C code.

The new tests not only check that the commands leave expected results in the cases where these new variables are set (with or without command line overrides), but also seem to have checks to see if the commands behave the same way as before unless the new configuration variables are used.

It is very understandable for any developers (including me) to want to demonstrate that their shiny new toys work as they specified, and writing the positive tests (i.e. "does the feature kick in when the user does what the manual says, and does it leave the expected result?") is a very good discipline to protect the new features from future breakages.

But at the same time, we (again, including me) tend to forget the importance about negative tests (e.g. "does the feature refrain from kicking in when the user does not do what the additional part of the manual says, iow, uses the traditional way of running the commands, and does it leave the expected result without the new feature's effect?") because (1) it is rather boring, and (2) we believe too much in ourselves' ability not to break things.

So I already am liking the series even before reading a single line of code. Makes me hope that the changes are done with the same carefulness as the tests ;-).