[PATCH] scripts: Translate profile2linkerlist.

[PATCH] scripts: Translate profile2linkerlist.

[Jorge Natz]

profile2linkerlist.pl is a script that takes the sorted output of the 
readprofile command and turns it into a list of C functions in a 
format for linker lists. Make the script more portable across POSIX 
systems. This script does not use any shell-specific features, only 
POSIX ones. A POSIX shell is present on all UNIX systems, while Perl 
is not as commonplace.

Signed-off-by: Jorge Natz <jorgenatzdev@gmail.com>
---
diff -uprN a/scripts/profile2linkerlist.pl b/scripts/profile2linkerlist.pl
--- a/scripts/profile2linkerlist.pl	2016-06-10 17:27:15.333317028 -0600
+++ b/scripts/profile2linkerlist.pl	1969-12-31 17:00:00.000000000 -0700
@@ -1,19 +0,0 @@
-#!/usr/bin/perl
-
-#
-# Takes a (sorted) output of readprofile and turns it into a list suitable for
-# linker scripts
-#
-# usage:
-#	 readprofile | sort -rn | perl profile2linkerlist.pl > functionlist
-#
-use strict;
-
-while (<>) {
-  my $line = $_;
-
-  $_ =~ /\W*[0-9]+\W*([a-zA-Z\_0-9]+)\W*[0-9]+/;
-
-  print "*(.text.$1)\n"
-      unless ($line =~ /unknown/) || ($line =~ /total/);
-}
diff -uprN a/scripts/profile2linkerlist.sh b/scripts/profile2linkerlist.sh
--- a/scripts/profile2linkerlist.sh	1969-12-31 17:00:00.000000000 -0700
+++ b/scripts/profile2linkerlist.sh	2016-03-24 17:26:58.044958415 -0600
@@ -0,0 +1,18 @@
+#!/bin/sh
+
+# Takes a (sorted) output of readprofile and turns it into a list suitable for
+# linker scripts.
+#
+# usage:
+#         readprofile | sort -rn | ./profile2linkerlist.sh > functionlist
+
+# Exit on error
+set -e
+
+while read LINE; do
+	LINE=$( echo $LINE | sed -e s"/[0-9.][0-9.]*//" -e s"/[0-9.][0-9.]*$//" | xargs)
+	case $LINE in *unknown*|*total* ) continue
+	;;
+	esac
+	echo '*(.text.'$LINE')'
+done

Re: [PATCH] scripts: Translate profile2linkerlist.

[Michal Marek]

Dne 7.8.2016 v 20:33 Jorge Natz napsal(a):
> -while (<>) {
> -  my $line = $_;
> -
> -  $_ =~ /\W*[0-9]+\W*([a-zA-Z\_0-9]+)\W*[0-9]+/;
> -
> -  print "*(.text.$1)\n"
[...]
> +	LINE=$( echo $LINE | sed -e s"/[0-9.][0-9.]*//" -e s"/[0-9.][0-9.]*$//" | xargs)
> +	case $LINE in *unknown*|*total* ) continue
> +	;;
> +	esac
> +	echo '*(.text.'$LINE')'
> +done

The two regular expressions are not equivalent and the shell version
does not take into account the locale-dependent decimal separator. It
also does not handle localized "total", but that's an issue of the perl
version as well. Since this script is not used during kernel build, its
dependency on perl is less of an issue and it can be left as is.

Michal