v0.36.0

The whole way

One hop was as far as it went, because what a program is given was one run of
octets and the first program used all of it, so the program it became had
nothing left to read. It is a sequence now.

the board       loads a kernel of 2.6 kilooctets
the kernel      loads a compiler of 137 kilooctets
the compiler    reads 57 kilooctets of its own language off the disk,
                compiles it to a front end of 192 kilooctets, hands it
                back, and stops being the compiler
the front end   reads the glyph's source off the disk, compiles it to
                1389 octets, hands it back, and stops being the front end
what is left    prints the figure

Two compilers, three programs, one machine, four tenths of a second.

One thing had to be fixed for it, and it is the sort a second hop finds and a
first does not: the routine that takes a program out of memory read the
headers out of the buffer it was handed, and a segment clearing sixteen
megaoctets of its own memory clears that buffer, with the entry point in it.

Full Changelog: v0.35.0...v0.36.0

LOC: 19906