[merge] knit index robustness/corruption?

[John Arbash Meinel]

John Arbash Meinel has voted +0.
Status is now: Semi-approved
Comment:
I'd rather still have an exception raised, because we know this is 
actual corruption. I won't stick on this.

Tighter exception handling would be preferred.

I think we *should* have a test for this. And I think we already have 
some knit index corruption tests. (At least for the "line is not 
properly terminated" case). It should be easy to do in the 
LowLevelKnitIndexTests, since we explicitly generate the knit index data 
there.)

So for a +1 you at least need a test. You can decide on the rest 
(warning/error, multiple try/except or just one)

For details, see: 
http://bundlebuggy.aaronbentley.com/request/%3Ce01316480706250224w43e18344n1e271e4a00e4074f%40mail.gmail.com%3E