Re: lines of code statistics

[Roland McGrath]

> I run the sloccount program on libc, gnumach, hurd, oskit-mach and mig.

Hey, fresh bullshit! :-)

> libc contains support for many platforms, one reason why it adds
> substantially to the overall size.
> gnumach is six times larger than oskit-mach in number of code lines.
> But don't forget that oskit is larger than gnumach ;)

There is a lot of dead code in gnumach that I removed in oskit-mach.  And I
bet the drivers are the majority of the LOC in any kernel (that has drivers
in it at all, which is why oskit-mach itself is so small).

Things like libc, and especially the oskit, have a lot of code that never
actually goes into any part of a GNU/Hurd system.  So these kinds of counts
are always pretty unrealistic.  (However, that is only appropriate as you
then multiply them by the utterly ludicrous development effort estimate
based solely on source language.)