Re: [util-linux] readprofile ignores the last element in /proc/profile

[William Lee Irwin III <wli@holomorphy.com>]

On Sun, Aug 29, 2004 at 02:23:50PM -0700, William Lee Irwin III wrote:
> It probably qualifies as impolite. I suppose a better way of going
> about this would be saying that the code has accumulated some cruft,
> and then describing the differences in greater detail.
> The removal of the multiplier resetting is an oversight; I rarely
> use the feature myself, but upon closer examination, the multiplier
> accepted by write_profile() is not ASCII. I also question the role of a
> profile report extraction utility in alterations of profiling state.
> The removal of the reset feature is once again intentional as this
> can be done by echo > /proc/profile.
> The removal of the individual bin count reporting is intentional, but
> not a very nice removal, as it's a useful feature and definitely not a
> misfeature. The format of the histogram reporting is, however, not very
> useful as sort(1) etc. don't easily interoperate with it. This should
> be put back for serious use.

Oh, the removal of the reversed byte order detection was also very
intentional as that appears to frequently go wrong. I'd prefer an
explicit option to support cross-endianness profiling.


-- wli