[Gllug] Bug in grep? Surely not ...

[Peter Grandi]

>>> On Tue, 20 Sep 2005 15:06:08 +0100, Richard Jones
>>> <rich at annexia.org> said:

rich> I have this rather large data file, and I want to get out
rich> just the non-zero rows: http://annexia.org/tmp/col4 [
rich> ... ] BUT if I redirect the output to a file, grep hangs
rich> consuming 100% CPU,

It is practically instantaneous here.

As a first step, check whether the 100% CPU consumption is user
or system CPU with 'vmstat 1'.

rich> I'm quite sure this can't _really_ be a bug in grep, but
rich> I'm at a loss to know where to look. [ ... ]

It is hard to imagine that it is a 'grep' because the output
file descriptor is set up by the shell, not 'grep', in either
case.

Perhaps something is wrong with the filesystem in which '/tmp'
is contained. So try to redirect output first to '/dev/null'
and then to a file on another filesystem/partition. Same for
input BTW, as that's on '/tmp' too.

As to my '/tmp', it is a 'tmpfs', but the same 'grep' is also
instantaneous here to '/var/tmp/costs', which is on my 'root'
filesystem.

rich> but strace shows its not doing anything at all.

Well, 'strace' only shows system calls -- if a program gets
stuck in a looping system call... :-)

You can also use GDB in attach mode to see exactly what the
'grep' is doing, or perhaps use 'ltrace' as an intermediate
detail solution.

-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug