aboutsummaryrefslogtreecommitdiff
path: root/perl/Curl_easy/t/01basic.t
diff options
context:
space:
mode:
Diffstat (limited to 'perl/Curl_easy/t/01basic.t')
-rw-r--r--perl/Curl_easy/t/01basic.t93
1 files changed, 0 insertions, 93 deletions
diff --git a/perl/Curl_easy/t/01basic.t b/perl/Curl_easy/t/01basic.t
deleted file mode 100644
index 79a6e8a24..000000000
--- a/perl/Curl_easy/t/01basic.t
+++ /dev/null
@@ -1,93 +0,0 @@
-# Test script for Perl extension Curl::easy.
-# Check out the file README for more info.
-
-# Before `make install' is performed this script should be runnable with
-# `make t/thisfile.t'. After `make install' it should work as `perl thisfile.t'
-
-######################### We start with some black magic to print on failure.
-
-# Change 1..1 below to 1..last_test_to_print .
-use strict;
-
-BEGIN { $| = 1; print "1..6\n"; }
-END {print "not ok 1\n" unless $::loaded;}
-use Curl::easy;
-
-$::loaded = 1;
-print "ok 1\n";
-
-######################### End of black magic.
-
-# Insert your test code below (better if it prints "ok 13"
-# (correspondingly "not ok 13") depending on the success of chunk 13
-# of the test code):
-
-my $count=1;
-
-# Read URL to get
-my $defurl = "http://localhost/cgi-bin/printenv";
-my $url;
-if (defined ($ENV{CURL_TEST_URL})) {
- $url=$ENV{CURL_TEST_URL};
-} else {
-$url = "";
-print "Please enter an URL to fetch [$defurl]: ";
-$url = <STDIN>;
-if ($url =~ /^\s*\n/) {
- $url = $defurl;
-}
-}
-
-# Init the curl session
-my $curl = Curl::easy::init();
-if ($curl == 0) {
- print "not ";
-}
-print "ok ".++$count."\n";
-
-Curl::easy::setopt($curl, CURLOPT_NOPROGRESS, 1);
-Curl::easy::setopt($curl, CURLOPT_MUTE, 1);
-Curl::easy::setopt($curl, CURLOPT_FOLLOWLOCATION, 1);
-Curl::easy::setopt($curl, CURLOPT_TIMEOUT, 30);
-
-open HEAD, ">head.out";
-Curl::easy::setopt($curl, CURLOPT_WRITEHEADER, *HEAD);
-print "ok ".++$count."\n";
-
-open BODY, ">body.out";
-Curl::easy::setopt($curl, CURLOPT_FILE,*BODY);
-print "ok ".++$count."\n";
-
-my $errbuf;
-Curl::easy::setopt($curl, CURLOPT_ERRORBUFFER, "errbuf");
-print "ok ".++$count."\n";
-
-Curl::easy::setopt($curl, CURLOPT_URL, $url);
-
-print "ok ".++$count."\n";
-# Add some additional headers to the http-request:
-my @myheaders;
-$myheaders[0] = "Server: www";
-$myheaders[1] = "User-Agent: Perl interface for libcURL";
-Curl::easy::setopt($curl, CURLOPT_HTTPHEADER, \@myheaders);
-
-my $bytes;
-my $realurl;
-my $httpcode;
-
-# Go get it
-my $retcode=Curl::easy::perform($curl);
-if ($retcode == 0) {
- Curl::easy::getinfo($curl, CURLINFO_SIZE_DOWNLOAD, $bytes);
- print STDERR "$bytes bytes read ";
- Curl::easy::getinfo($curl, CURLINFO_EFFECTIVE_URL, $realurl);
- Curl::easy::getinfo($curl, CURLINFO_HTTP_CODE, $httpcode);
- print STDERR "effective fetched url (http code: $httpcode) was: $url ";
-} else {
- # We can acces the error message in $errbuf here
- print STDERR "$retcode / '$errbuf'\n";
- print "not ";
-}
-print "ok ".++$count."\n";
-
-exit;