[Gnu-arch-users] into the ether again: wd factor

[Robert Anderson]

I'll throw this one into the ether again:

A patch based rcs like arch would be far more useful if people actually 
made logically distinct, clean patches with it.  This is _much_ harder 
than it should be.  There's no tools for it.  There really ought to be.

The work pattern of "think of clean, logically distinct feature, code 
it, commit it" doesn't exist in the wild, at least not in my world.  
Closer to reality is: start working, realize you really need to factor 
out a helper function first, return to working, clean up some ugliness 
in unrelated code that you happened to scroll by, return to working, 
implement a bunch of debug/diagnostic junk that you'll want out later, 
return to working, repeat various loosely related tasks until you really 
shouldn't go any further without committing if you have any remaining 
self respect.

Now you've got a working directly that represents what really ought to 
be four or five changesets, but there's no good way to split them out 
without a hell of a lot of careful work, so you commit the whole ball of 
wax and flog yourself for not being more disciplined in the first 
place.  But hindsight is always 20/20 and you've got to keep moving 
forward.  You can't keep stopping your workflow to consider if every 
change you make should be predicated by mucking with your rcs.  This 
kind of factoring into clean, logical changesets has to be an 
aposteriori process in the real world.  I'm fully convinced of that.

Then two weeks later you realize you want all that debug junk out, and 
wish it was its own patch so you could reverse it and move on with your 
life.  Or you realize there's a bug and want to isolate it among the 
various orthogonal changes all living in that one patch.  Too bad.  
You've got changesets, but you're not really exploiting their full 
potential.

If there was a way to take a tree and split out changes hunk by hunk 
into two or more trees (including a null tree), wouldn't the power of 
changesets really finally be there in a way that's far more useful?

This would be an interactive tool, of course, so it's not an arch 
command, but more the domain of possibly one of the emacs mode writers 
or one of the wrapper writers.

I'm just really surprised that this hasn't already been done, redone, 
and perfected years ago, it seems that fundamental to me.

Bob