public inbox for linux-ia64@vger.kernel.org
 help / color / mirror / Atom feed
* making unwcheck.sh
@ 2003-11-14 18:38 David Mosberger
  2003-11-15  6:20 ` Matt Chapman
  0 siblings, 1 reply; 3+ messages in thread
From: David Mosberger @ 2003-11-14 18:38 UTC (permalink / raw)
  To: linux-ia64

Is anybody willing/capable/interested in fixing unwcheck.sh so it can
actually handle 64-bit addresses?  I'm not much of an awk hacker but I
noticed that the only reason the script works on the kernel is because
it drops the first digit of each address (look for gsub() to see what
I mean).  Of course, that makes it work on the kernel, but it won't
work for checking a shared object, for example.  Perhaps the whole
script should just be converted to a real[1] language?


	--david

[1] real = anything other than awk, perl, or tcl... ;-)

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: making unwcheck.sh
  2003-11-14 18:38 making unwcheck.sh David Mosberger
@ 2003-11-15  6:20 ` Matt Chapman
  0 siblings, 0 replies; 3+ messages in thread
From: Matt Chapman @ 2003-11-15  6:20 UTC (permalink / raw)
  To: linux-ia64

What's wrong with Perl? :)  Okay, so there's plenty wrong with Perl, but
I think it's a decent tool for the job.

I wrote a quick Perl version, though it only works properly on 64-bit
platforms (reports an integer overflow otherwise).  It wouldn't be too
hard to work around this if it's an issue (since for a given function
the upper 32 bits should presumably be invariant), in fact the same
workaround could probably be done in the awk script.

http://www.cse.unsw.edu.au/~matthewc/files/unwcheck.pl

Python seems to handle big numbers better but I don't really know enough
Python.

Matt


On Fri, Nov 14, 2003 at 10:38:39AM -0800, David Mosberger wrote:
> Is anybody willing/capable/interested in fixing unwcheck.sh so it can
> actually handle 64-bit addresses?  I'm not much of an awk hacker but I
> noticed that the only reason the script works on the kernel is because
> it drops the first digit of each address (look for gsub() to see what
> I mean).  Of course, that makes it work on the kernel, but it won't
> work for checking a shared object, for example.  Perhaps the whole
> script should just be converted to a real[1] language?
> 
> 
> 	--david
> 
> [1] real = anything other than awk, perl, or tcl... ;-)
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ia64" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: making unwcheck.sh
@ 2004-01-15 23:33 David Mosberger
  0 siblings, 0 replies; 3+ messages in thread
From: David Mosberger @ 2004-01-15 23:33 UTC (permalink / raw)
  To: linux-ia64

>>>>> On Sat, 15 Nov 2003 17:20:48 +1100, Matt Chapman <matthewc@cse.unsw.edu.au> said:

  Matt> Python seems to handle big numbers better but I don't really
  Matt> know enough Python.

Well, attached is what I came up with (I'm not terribly proficient
in Python, I should say, so suggestions for improvements are welcome).

Of course, it promptly found a problem in mca_asm.S:

 ERROR: function ia64_monarch_init_handler: 186 slots, sum of region len = 179

Actually, the entire mca_asm.S file looks rather atrocious.  For
example, it doesn't take advantage of the ENTRY/END macros in
asm_macro.h.  Could somebody look into fixing and cleaning this up?

Thanks,

	--david

#!/usr/bin/env python
#
# Usage: unwcheck.py FILE
#
# This script checks the unwind info of each function in file FILE
# and verifies that the sum of the region-lengths matches the total
# length of the function.
#
# Based on a shell/awk script originally written by Harish Patil,
# which was converted to Perl by Matthew Chapman, which was converted
# to Python by David Mosberger.
#
import os
import re
import sys

if len(sys.argv) != 2:
    print "Usage: %s FILE" % sys.argv[0]
    sys.exit(-1)

readelf = os.getenv("READELF", "readelf")

start_pattern = re.compile("<([^>]*)>: \[0x([0-9a-f]+)-0x([0-9a-f]+)\]")
rlen_pattern  = re.compile(".*rlen=([0-9]+)")

def check_func (func, slots, rlen_sum):
    if slots != rlen_sum:
        global num_errors
        num_errors += 1
        if not func: func = "at %#x" % start
        print "ERROR: function %s: %lu slots, sum of region lengths = %lu" \
              % (func, slots, rlen_sum)
    return

num_funcs = 0
num_errors = 0
func = False
slots = 0
rlen_sum = 0
for line in os.popen("%s -u %s" % (readelf, sys.argv[1])):
    m = start_pattern.match(line)
    if m:
        check_func(func, slots, rlen_sum)

        func  = m.group(1)
        start = long(m.group(2), 16)
        end   = long(m.group(3), 16)
        slots = 3 * (end - start) / 16
        rlen_sum = 0L
        num_funcs += 1
    else:
        m = rlen_pattern.match(line)
        if m:
            rlen_sum += long(m.group(1))
check_func(func, slots, rlen_sum)

if num_errors = 0:
    print "No errors detected in %u functions." % num_funcs
else:
    print "%u errors detected in %u functions." % (num_errors, num_funcs)
    sys.exit(-1)

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2004-01-15 23:33 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-11-14 18:38 making unwcheck.sh David Mosberger
2003-11-15  6:20 ` Matt Chapman
  -- strict thread matches above, loose matches on Subject: below --
2004-01-15 23:33 David Mosberger

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox