grep MBS_SUPPORT is a dog

[Neal Nuckolls]

System:
maytag$ uname -a
Linux maytag 2.4.20-8 #1 Thu Mar 13 17:54:28 EST 2003 i686 i686 i386 GNU/Linux
maytag$ cat /etc/redhat-release
Red Hat Linux release 9 (Shrike)
maytag$ rpm -q -f /bin/grep
grep-2.5.1-7

Description:
grep takes almost 2 seconds realtime when matching one or more lines
in a small /tmp textfile .  When no lines are matched, it ran in
only a few milliseconds (reasonable).  But when one or more lines matched,
it took over 1.8 *seconds* (not reasonable).  The small text file
should easily be cached in-memory page cache (no i/o) .

Here's the file:

maytag$ wc -l /tmp/file
   1832 /tmp/file
maytag$ ls -l /tmp/file
-rw-------    1 nn       users      189220 Nov  2 09:51 /tmp/file

Here's the time for nomatch (ok):

maytag$ time /bin/grep -c jffjf /tmp/file
0

real    0m0.078s
user    0m0.070s
sys     0m0.010s
maytag$ time /bin/grep -c jffjf /tmp/file
0

real    0m0.077s
user    0m0.070s
sys     0m0.000s


Here's the time for match (not ok):

maytag$ time /bin/grep -c suspended /tmp/file
336

real    0m1.800s
user    0m1.790s
sys     0m0.010s
maytag$ time /bin/grep -c suspended /tmp/file
336

real    0m1.800s
user    0m1.790s
sys     0m0.010s

---

Downloading grep-2.5.1-7.8.src.rpm, compiling, and gprof'ing shows
100% of the time is in check_multibyte_string() .
I don't care a #*$% about multibyte character support.
Editing search.c and dfa.c to remove the "#define MBS_SUPPORT"
and the new grep performs reasonably:

maytag$ time /tmp/grep -c suspended /tmp/file
336

real    0m0.009s
user    0m0.010s
sys     0m0.000s
maytag$ time /tmp/grep -c suspended /tmp/file
336

real    0m0.009s
user    0m0.000s
sys     0m0.010s


MBS_SUPPORT is a dog.  Please either fix it or disable it by default.  sheesh.


neal nuckolls
address@hidden