[Toybox] Release fiddling, need another day or two.

[Rob Landley]

I have rather a lot of open email windows with 2/3 finished replies in
them....

On 10/04/2017 04:25 PM, scsijon wrote:
>> So I'm taking a little extra time to finish off at least one, hopefully
>> two. Stalling on the way to 1.0 is not cool and I'd like to get
>> unblocked.
>>
>> Rob
>>
> Instead, think of it as a "getting most of the bugs out and having a
> working base point" release.

I had that already (and eventually shipped it), but I also have new code
in xargs that doesn't have a corresponding test...

And writing that test code I found out lifting the _total_ arg+env space
restriction in the kernel means the the max length of an _individual_
argument is still 131071. (Which we don't need to test in xargs because
what do you do about it, we can't police the individual inputs, all we
could do is skip them which isn't what we were asked to do.) But it
means that testing the length limit edge case means I can't just do
"printf '%0*d' $BIGVALUE 0 | xargs /bin/echo", I have to split it up
into blocks of 131071 plus the tail.

Grinding through my todo list to keep up with the new items it spins off
_as_ I grind through it would be a full time job, and the past 18 months
or so I've _had_ a full time job that's actually requiring all the
energy I can spare for it...

> Well worth having from time to time
> Especially if later problems cause things suddenly turn to quicksand,
> you can check if the problem cause was in code before or after this
> point, and that shortens rework cycles to get up and running again.

That's not a release question, that's a "when do I commit" question. I
try very hard to test that sort of thing _before_ it goes into the tree,
at least now that there are users. Of course that leads to:

$ git diff | wc
   1797    8542   63136

And that's _after_ I just cut a release.

> Anyway, keep up the good work, I DO like toybox.

Thanks.

> regards
> scsijon
> ps I wonder if android could make use of it?

I believe Elliott pulls my tree on fridays, and they use their own build
system so my releases are largely cosmetic from their point of view. I
think M, N, and O all shipped git snapshot du jour when they froze,
rather than release versions. But they do their own testing, and I try
to test each commit, so it works out.

(Releases are good because they provide coordination points, but android
has its own releases which are the coordination points they care about.)

Rob