Re: Improved ksymoops output

[Roberto Nibali <ratz@drugphish.ch>]

Hello,

I like the idea very much. As I have to maintain a kernel tree for 
various customers too, I think this could be a nice addition for the bug 
reports (seldom luckily) I sometimes get.

> gen_func2file.map
> =================

[snipped rest of the bash script]

This can be improved a little bit:

----------------------------------------------------------------
#!/bin/sh
# Meant to be run in top lever kernel source dir
# after kernel has been built
#
# Makes list of the form:
# symbol1 object_file_pathname1
# symbol2 object_file_pathname2
# symbol3 object_file_pathname3

 > func2file.map

find -name '*.c' | while read a; do
     #nm: get symbols from .o
     #grep: discard non-text symbols
     #awk: remove './', add .o pathname
     #cut: remove address and symbol type letter
     b=${a%.*}
     test -e "${b}.o" && {
         nm "${b}.o" \
         | grep '\( T \)\|\( t \)' \
         | awk "BEGIN { N=\" ${a:2}\" } { print \$0,N }" \
         | cut -b12- \
         >> func2file.map
     }
done
----------------------------------------------------------------

This script compared to yours does (it's nitpicking):
o run faster (5%) ;)
o should never have problems when one day there will be a lot of *.c
   files. In your approach LIST could someday not hold all entries
   anymore.
o simplifies the bash 'regexp' to snip away the '.c' and print the rest

I'm propably going to rewrite the python script in bash too, since I 
don't run python on my distro (and I do not intend to use 2.5.x anytime 
soon).

Best regards,
Roberto Nibali, ratz