Re: slow grep

[Stepan Kasal]

Hello,
        long time ago, you've reported the following bug:

On Wed, Nov 13, 2002 at 03:50:00PM -0800, Aaron Holtzman wrote:
> [posted to both glibc and gnu-utils as the problem may be in either]
> 
> Using grep-2.5.1 / glibc-2.2.93-5 (RedHat 8) I see a very large
> performance degradation for the following search pattern.
> 
>       wilder:/tmp$ time grep  -E '  +[0-9]+' test_input  > /dev/null
> 
>       real    0m2.334s
>       user    0m2.330s
>       sys     0m0.002s
> 
> For reference, the same test done on another (slower) machine, with
> grep-2.4.2 and glibc-2.2.4-30 (RedHat 7.3).
> 
>       kurosawa:/tmp$ time grep -E  '  +[0-9]+' test_input  > /dev/null
> 
>       real    0m0.004s
>       user    0m0.000s
>       sys     0m0.000s
> 
> I've attached the test_input file for your convenience.

RedHat 8 uses Unicode by default. For example it has LANG="en_GB.UTF-8"
in situations where RedHat 7.3 used to have LANG="en_GB.iso885915".

This was probably the cause of the problem.

I'd suggest to go back to the old setting or use LANG=C for now.

Please forgive us that it took so long to produce this answer.

Stepan Kasal