curl/tests/keywords.pl

154 строки
3.3 KiB
Perl
Executable File

#!/usr/bin/env perl
#***************************************************************************
# _ _ ____ _
# Project ___| | | | _ \| |
# / __| | | | |_) | |
# | (__| |_| | _ <| |___
# \___|\___/|_| \_\_____|
#
# Copyright (C) 1998 - 2014, Daniel Stenberg, <daniel@haxx.se>, et al.
#
# This software is licensed as described in the file COPYING, which
# you should have received as part of this distribution. The terms
# are also available at http://curl.haxx.se/docs/copyright.html.
#
# You may opt to use, copy, modify, merge, publish, distribute and/or sell
# copies of the Software, and permit persons to whom the Software is
# furnished to do so, under the terms of the COPYING file.
#
# This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
# KIND, either express or implied.
#
###########################################################################
use strict;
push(@INC, $ENV{'srcdir'}) if(defined $ENV{'srcdir'});
push(@INC, ".");
require "getpart.pm"; # array functions
my $srcdir = $ENV{'srcdir'} || '.';
my $TESTDIR="$srcdir/data";
# Get all commands and find out their test numbers
opendir(DIR, $TESTDIR) || die "can't opendir $TESTDIR: $!";
my @cmds = grep { /^test([0-9]+)$/ && -f "$TESTDIR/$_" } readdir(DIR);
closedir DIR;
my $TESTCASES; # start with no test cases
# cut off everything but the digits
for(@cmds) {
$_ =~ s/[a-z\/\.]*//g;
}
# the the numbers from low to high
for(sort { $a <=> $b } @cmds) {
$TESTCASES .= " $_";
}
my $t;
my %k; # keyword count
my %t; # keyword to test case mapping
my @miss; # test cases without keywords set
my $count;
my %errors;
for $t (split(/ /, $TESTCASES)) {
if(loadtest("${TESTDIR}/test${t}")) {
# bad case
next;
}
my @ec = getpart("verify", "errorcode");
if($ec[0]) {
# count number of check error codes
$errors{ 0 + $ec[0] } ++;
}
my @what = getpart("info", "keywords");
if(!$what[0]) {
push @miss, $t;
next;
}
for(@what) {
chomp;
#print "Test $t: $_\n";
$k{$_}++;
$t{$_} .= "$t ";
}
$count++;
}
sub show {
my ($list)=@_;
my @a = split(" ", $list);
my $ret;
my $c;
my @l = sort {rand(100) - 50} @a;
my @ll;
for(1 .. 11) {
my $v = shift @l;
if($v) {
push @ll, $v;
}
}
for (sort {$a <=> $b} @ll) {
if($c++ == 10) {
$ret .= "...";
last;
}
$ret .= "$_ ";
}
return $ret;
}
# sort alphabetically
my @mtest = reverse sort { lc($b) cmp lc($a) } keys %k;
print <<TOP
<table><tr><th>Num</th><th>Keyword</th><th>Test Cases</th></tr>
TOP
;
for $t (@mtest) {
printf "<tr><td>%d</td><td>$t</td><td>%s</td></tr>\n", $k{$t},
show($t{$t});
}
printf "</table><p> $count out of %d tests (%d lack keywords)\n",
scalar(@miss) + $count,
scalar(@miss);
for(@miss) {
print "$_ ";
}
print "\n";
printf "<p> %d different error codes tested for:<br>\n",
scalar(keys %errors);
# numerically on amount, or alphebetically if same amount
my @etest = sort { $a <=> $b} keys %errors;
for(@etest) {
print "$_ ";
}
print "\n";