# watch the end of the current debug file (like tail -f) applying
# any regexes supplied on the command line.
#
+# There can be more than one <regexp>. a <regexp> preceeded by a '!' is
+# treated as NOT <regexp>. Each <regexp> is implcitly ANDed together.
+# All <regexp> are caseless.
+#
# examples:-
#
# watchdbg g1tlh # watch everything g1tlh does
my $today = $fp->unixtoj(time());
my $fh = $fp->open($today) or die $!;
my $nolines = 1;
-$nolines = shift if $ARGV[0] =~ /^-?\d+$/;
-$nolines = abs $nolines if $nolines < 0;
-my $exp = join '|', @ARGV;
+my @patt;
my @prev;
+while (@ARGV) {
+ my $arg = shift;
+ if ($arg =~ /^-+(\d+)/) {
+ $nolines += $1;
+ next;
+ }
+ usage(), exit(0) if $arg =~ /^-+[h\?]/i;
+ push @patt, $arg;
+}
+
+
# seek to end of file
$fh->seek(0, 2);
for (;;) {
my $line = $fh->getline;
if ($line) {
- if ($exp) {
+ if (@patt) {
push @prev, $line;
shift @prev while @prev > $nolines;
- if ($line =~ m{(?:$exp)}oi) {
+ my $flag = 0;
+ foreach my $p (@patt) {
+ if ($p =~ /^!/) {
+ my $r = substr $p, 1;
+ last if $line =~ m{$r}i;
+ } else {
+ last unless $line =~ m{$p}i;
+ }
+ ++$flag;
+ }
+ if ($flag == @patt) {
printit(@prev);
@prev = ();
}
$line =~ s/([\x00-\x1f\x7f-\xff])/sprintf("\\x%02X", ord($1))/eg;
my ($t, $l) = split /\^/, $line, 2;
$t = time unless defined $t;
- printf "%02d:%02d:%02d %s\n", (gmtime($t))[2,1,0], $l;
+ printf "%02d:%02d:%02d %s\n", (gmtime($t))[2,1,0], $l;
}
}
exit(0);
+
+sub usage
+{
+ print << "XXX";
+
+ usage: watchdbg [-nnn lines before] [<regexp>|!<regexp>]...
+
+ You can have more than one <regexp> with an implicit 'and' between them. All
+ <regexes> are caseless. It's recommended to put 'not' (!<regex>) first in any list.
+ Don't forget that you are doing this in a shell and you may need to quote your
+ <regex>s.
+
+ watchdbg -2 progress
+
+ will display any line containing 'progress' and also the two lines before that.
+
+XXX
+}