lkml.org 
[lkml]   [2010]   [Feb]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Subject[PATCH] scripts/checkpatch.pl: Dramatically improve #define parse times
From
Date
#define macros currently take a _very_ long time to parse because
the macro is being searched for multiple statements and unnecessary
searching is being done when comments are replaced with unprintable
characters.

For instance:

Without this patch:
$ time perl ./scripts/checkpatch.pl -f include/linux/mfd/wm831x/irq.h > /dev/null

real 3m43.229s
user 3m42.254s
sys 0m0.332s

With this patch:
$ time perl ./scripts/checkpatch.pl -f include/linux/mfd/wm831x/irq.h > /dev/null

real 0m1.372s
user 0m1.348s
sys 0m0.020s

A 160:1 improvement.

I don't know if this is correct, but it seems to work for me.

Maybe there's some reason why C comments aren't replaced with
spaces but get unprintable characters.

Changes:

Replace comments with spaces.
Do not call ctx_statement_block unless necessary
Add a check to see if the macro could have multiple statements
either with a \ line continuation or a ";". This doesn't work with
comma separated statements.
Miscellaneous its -> it's typo corrections.

Signed-off-by: Joe Perches <joe@perches.com>
---
diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index 3257d3d..70747c8 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -373,18 +373,18 @@ sub sanitise_line {
for ($off = 1; $off < length($line); $off++) {
$c = substr($line, $off, 1);

- # Comments we are wacking completly including the begin
+ # Comments we are whacking completely including the begin
# and end, all to $;.
if ($sanitise_quote eq '' && substr($line, $off, 2) eq '/*') {
$sanitise_quote = '*/';

- substr($res, $off, 2, "$;$;");
+ substr($res, $off, 2, " ");
$off++;
next;
}
if ($sanitise_quote eq '*/' && substr($line, $off, 2) eq '*/') {
$sanitise_quote = '';
- substr($res, $off, 2, "$;$;");
+ substr($res, $off, 2, " ");
$off++;
next;
}
@@ -417,9 +417,9 @@ sub sanitise_line {

#print "c<$c> SQ<$sanitise_quote>\n";
if ($off != 0 && $sanitise_quote eq '*/' && $c ne "\t") {
- substr($res, $off, 1, $;);
+ substr($res, $off, 1, ' ');
} elsif ($off != 0 && $sanitise_quote eq '//' && $c ne "\t") {
- substr($res, $off, 1, $;);
+ substr($res, $off, 1, ' ');
} elsif ($off != 0 && $sanitise_quote && $c ne "\t") {
substr($res, $off, 1, 'X');
} else {
@@ -507,7 +507,7 @@ sub ctx_statement_block {
last;
}

- # An else is really a conditional as long as its not else if
+ # An else is really a conditional as long as it's not else if
if ($level == 0 && $coff_set == 0 &&
(!defined($p) || $p =~ /(?:\s|\}|\+)/) &&
$remainder =~ /^(else)(?:\s|{)/ &&
@@ -1432,7 +1432,9 @@ sub process {
# Check for potential 'bare' types
my ($stat, $cond, $line_nr_next, $remain_next, $off_next,
$realline_next);
- if ($realcnt && $line =~ /.\s*\S/) {
+ my $test = $line =~ /\+\s*(\S)/;
+ my $firstchar = $1;
+ if ($realcnt && $test && $firstchar ne "\#") {
($stat, $cond, $line_nr_next, $remain_next, $off_next) =
ctx_statement_block($linenr, $realcnt, 0);
$stat =~ s/\n./\n /g;
@@ -1880,7 +1882,7 @@ sub process {
# cpp #elif statement condition may start with a (
} elsif ($ctx =~ /^.\s*\#\s*elif\s*$/) {

- # If this whole things ends with a type its most
+ # If this whole things ends with a type it's most
# likely a typedef for a function.
} elsif ($ctx =~ /$Type$/) {

@@ -2280,10 +2282,11 @@ sub process {
}

# multi-statement macros should be enclosed in a do while loop, grab the
-# first statement and ensure its the whole macro if its not enclosed
+# first statement and ensure it's the whole macro if it's not enclosed
# in a known good container
if ($realfile !~ m@/vmlinux.lds.h$@ &&
- $line =~ /^.\s*\#\s*define\s*$Ident(\()?/) {
+ $line =~ /^.\s*\#\s*define\s*$Ident(\()?/ &&
+ ($line =~ /;/ || $line =~ /\\$/)) {
my $ln = $linenr;
my $cnt = $realcnt;
my ($off, $dstat, $dcond, $rest);



\
 
 \ /
  Last update: 2010-02-03 04:21    [W:0.106 / U:5.296 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site