On 21.01.2013 13:14, Colin Watson wrote:

> Part seven.  This particular patch in the set loses some of the
> compressed-kernel size we'd gained from earlier patches, but it's still
> a good bit smaller than when I started on this work so I'm not worried.
> Note that, as a general pattern, I have intentionally not zeroed out
> read_hook_data after calling code using a read hook; this is because (a)
> the code that calls read hooks only checks whether read_hook is NULL,
> (b) every function that sets read_hook to something non-NULL naturally
> also sets read_hook_data if it cares about its value, and (c) avoiding
> the extra code to zero read_hook_data makes a few bytes of difference to
> the size of compressed core images.


There are some possible improvements that we've discussed on IRC.
Apparently you didn't have time to implement them. They can come in
separately later if needed. Go ahead.