[RFC,0/5] builtin/grep.c: fix a tiny logic flaw

Message

Rasmus Villemoes Feb. 12, 2019, 10:26 p.m. UTC

Background: I noticed that the condition in add_work() for when the
producer could add a new item was oddly different from the other
conditions on the todo_* bookkeeping variables - namely, in the other
cases we want todo_a != todo_b, whereas in add_work the condition is
todo_a+1!=todo_b. Another hint that something is slightly off is that
the code would break down if TODO_SIZE was set to 1.

The practical effect is negligible, and fixing it seems to be a bit
involved, hence probably not worth the churn - and if that's the
verdict, I suggest adding a comment in add_work() for future readers
and/or people who copy the producer/consumer logic to their own code.

Rasmus Villemoes (5):
  builtin/grep.c: change todo_* variables to unsigned
  builtin/grep.c: refactor loop in work_done() slightly
  builtin/grep.c: add shorthand for &todo[todo_end] in add_work()
  builtin/grep.c: add todo_item helper
  builtin/grep.c: fix fence-post error in add_work()

 builtin/grep.c | 40 ++++++++++++++++++++++++----------------
 1 file changed, 24 insertions(+), 16 deletions(-)