Changes touching this path

• the source-walk consumer census sees a helper brought in by a use, and which methods the Admitted exceptions census watches is read off the marker on their own docs: the consumer walk keyed a reach on a call site spelled with the module path, so the spawn-seam census in main.rs, which calls the cut through an import, was counted by nobody and stood on no discrimination floor at all, and the src file walk #1944 added to flags.rs was invisible the same way, which is why #1944 reported the consumer count unchanged. a use item is resolved per file now, with a rename, a glob and an item that does not close on its line refused rather than read wrongly, and the floor is owed by the consumers that read the cut rather than by every consumer, derived from this module own source as the helpers that reach MARKER, the one spelling the cut keys on, so the file walks and the header readers owe it nothing while every census that cuts owes it as before. the derived header sentence is 15 consumers in 3 compilation units: 9 library, 5 binary, 1 integration test, and the paragraph that enumerated the censuses by name now says what a consumer is and leaves the files as an index the guard checks one way, saying which way. with_leading_word, which rewrites argv by position with no spec asked, was documented as doing so and was watched by nothing, so it carries the marker now and arrives in the roster with its one caller, bisect mark_by_term; whole-word matching moves into source_walk beside the walk, one spelling where flags.rs held the copy. red under mutation, counts read each time with the filter selecting the one test: the imported half of the reach dropped (0 passed and 1 failed), a planted use-imported bare call of the cut in main.rs (0 passed and 1 failed), the floor call removed from the spawn census (0 passed and 1 failed), the cut-reader derivation reduced to the direct namers of MARKER (0 passed and 1 failed), a renamed import planted (0 passed and 1 failed), the marker taken off with_leading_word (0 passed and 1 failed), and a planted with_leading_word caller in bisect dispatch (0 passed and 1 failed). item 3 of the ticket was already closed by #1974 and nothing there moved. no migration, no wire or format byte moves and nothing outside a test and a doc comment moves, so this owes no deploy. the workspace suite is green (4013 passed over 125 binaries, 7 ignored) (#1946) 47b0d164 · dbf3dbe6…

Renames are not followed. loot's tree maps a path to an address, so a rename is a delete and an add. This list is the history of the name, not of the bytes.