Confusing warnings due to optimization
tavianator opened this issue · 2 comments
tavianator commented
Example 1
$ bfs -type f -type d
bfs: warning: bfs -type f -type d
bfs: warning: ~~~~~~~
bfs: warning: This expression is always false.
This part is good, but...
bfs: warning: bfs -type f -type d
bfs: warning: ~~~~~~~
bfs: warning: The result of this expression is ignored.
Why? Well, the optimizer sees bfs -type f -false at this point. Maybe this warning should only be printed if the -false was not the result of other optimizations (but how to track that?).
Example 2
$ bfs -not \( -type b -print \) -not \( -type c -print \)
bfs: warning: bfs -not ( -type b -print ) -not ( -type c -print )
bfs: warning: ~~~~ ~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~
bfs: warning: The result of this expression is ignored.
Here we De Morgan'd the expression into -not \( -type b -print -o -type c -print \), and we warn when stripping the unnecessary root -not. We picked the first -not as the span pretty arbitrarily. I'm not convinced this is a very useful warning even in the best case, actually:
$ bfs -not -print
bfs: warning: bfs -not -print
bfs: warning: ~~~~~~~~~~~
bfs: warning: The result of this expression is ignored.
tavianator commented
Example 3
Expression reordering leads to surprises:
$ bfs -not -type f -and -type p
bfs: warning: bfs -not -type f -and -type p
bfs: warning: ~~~~~~~
bfs: warning: This expression is always false.
tavianator commented
Example 4
"Unreachable" flags?
$ bfs -quit -O3
bfs: warning: bfs -quit -O3
bfs: warning: ~~~
bfs: warning: This expression is unreachable.