Place a temporary bandaid on the ls/du/df human-readable issue. This method is

not going to scale up as well as I would like, and Matt Kraai and I have
discussed a better long term solution.  But for now this will at least make all
the human-readable apps give correct answers.

Please test the human readable/non-human readable options on your systems!!!
 -Erik

1 files changed, 9 insertions, 6 deletions

diff --git a/ls.c b/ls.c
index 86691fd..49470e9 100644
--- a/ls.c
+++ b/ls.c
@@ -648,15 +648,18 @@ static int list_single(struct dnode *dn)
 					printf("%4d, %3d ", (int)MAJOR(dn->dstat.st_rdev), (int)MINOR(dn->dstat.st_rdev));
 				} else {
 #ifdef BB_FEATURE_HUMAN_READABLE
-					fprintf(stdout, "%9s ", make_human_readable_str(dn->dstat.st_size>>10,
-								(ls_disp_hr==TRUE)? 0: 1));
-#else
+					if (ls_disp_hr==TRUE) {
+						fprintf(stdout, "%9s ", make_human_readable_str(
+									dn->dstat.st_size>>10, 0));
+					} else 
+#endif	
+					{
 #if _FILE_OFFSET_BITS == 64
-					printf("%9lld ", dn->dstat.st_size);
+						printf("%9lld ", dn->dstat.st_size);
 #else
-					printf("%9ld ", dn->dstat.st_size);
-#endif
+						printf("%9ld ", dn->dstat.st_size);
 #endif
+					}
 				}
 				column += 10;
 				break;