Fix POSIX non-compliance in cfr(1)

Which slows it down a *lot*; add an issue.

2 files changed, 7 insertions, 2 deletions

diff --git a/ISSUES.markdown b/ISSUES.markdown
index 2715bfbb..27e28a0d 100644
--- a/ISSUES.markdown
+++ b/ISSUES.markdown
@@ -30,3 +30,6 @@ Known issues
 *   sxhkd(1) might be nicer than xbindkeys; it's in Debian Testing now
 *   Maybe I should port some of the prompt functions to POSIX sh so I don't
     have to maintain parallel versions for bash, pdksh, and zsh
+*   find(1) -printf isn't POSIX, and isn't even in OpenBSD's version. Is there
+    a nicer way to do cfr(1df) efficiently while still handling edge cases like
+    files with newlines? -exec printf . works, but it's sloooow ...
diff --git a/bin/cfr b/bin/cfr
index 9bb0e3bb..36c5a7cf 100755
--- a/bin/cfr
+++ b/bin/cfr
@@ -1,4 +1,6 @@
 #!/bin/sh
 # Count files recursively with find(1); this includes the given directories
-# themselves and their subdirectories as entries
-find "${@:-"$PWD"}" -printf . | wc -c
+# themselves and their subdirectories as entries, and correctly handles files
+# with newlines in their names. It's also all POSIX ... but on big directories,
+# it's *slow*.
+find "${@:-.}" -exec printf %.sx {} \; | wc -c