From: Adam Dickmeiss <adam@indexdata.dk>
Date: Tue, 13 Nov 2001 11:17:26 +0000 (+0000)
Subject: MIME check when reading HTTP header (not when reading content).
X-Git-Tag: ZMBOT.0.1~8
X-Git-Url: http://sru.miketaylor.org.uk/cgi-bin?a=commitdiff_plain;h=7476a63e6732f7f51eea10bf38daaea4a31be57f;p=tclrobot.git

MIME check when reading HTTP header (not when reading content).
File robots.txt always read - even when text/plain is denied.
---

diff --git a/robot.tcl b/robot.tcl
index f94c8d3..ee70b9a 100755
--- a/robot.tcl
+++ b/robot.tcl
@@ -1,5 +1,5 @@
 #!/usr/bin/tclsh 
-# $Id: robot.tcl,v 1.27 2001/11/09 13:26:50 adam Exp $
+# $Id: robot.tcl,v 1.28 2001/11/13 11:17:26 adam Exp $
 #
 proc RobotFileNext1 {area lead} {
     # puts "RobotFileNext1 area=$area lead=$lead"
@@ -50,7 +50,9 @@ proc RobotReadRecord {inf fromurlx distancex} {
 }
 
 proc RobotFileNext {area} {
-    global robotSeq global idletime ns
+    global robotSeq
+    global idletime ns
+    global status
 
     # puts "RobotFileNext robotSeq=$robotSeq"
     if {$robotSeq < 0} {
@@ -67,7 +69,7 @@ proc RobotFileNext {area} {
     if {![string length $n]} {
 	set robotSeq -1
 	flush stdout
-	puts "Round robin"
+	puts "Round robin un,ba,vi=$status(unvisited),$status(bad),$status(visited)"
 	return wait
     }
     incr robotSeq
@@ -99,6 +101,7 @@ proc RobotFileExist {area host path} {
 }
 
 proc RobotFileUnlink {area host path} {
+    global status
     # puts "RobotFileUnlink begin"
     # puts "area=$area host=$host path=$path"
     set lpath [split $path /]
@@ -109,10 +112,12 @@ proc RobotFileUnlink {area host path} {
     set npath $area/$host[join [lrange $lpath 0 $l] /d]/f$t
     # puts "npath=$npath"
     set comp [split $npath /]
+    if {[catch {exec rm [join $comp /]}]} return
+
     set l [llength $comp]
     incr l -1
-    if {[catch {exec rm [join $comp /]}]} return
     incr l -1
+    incr status($area) -1
     for {set i $l} {$i > 0} {incr i -1} {
         set path [join [lrange $comp 0 $i] /]
 	if {![catch {glob $path/*}]} return
@@ -130,6 +135,7 @@ proc RobotFileClose {out} {
 proc RobotFileOpen {area host path {mode w}} {
     set orgPwd [pwd]
     global workdir
+    global status
 
     if {![info exists workdir]} {
 	return stdout
@@ -157,6 +163,7 @@ proc RobotFileOpen {area host path {mode w}} {
 		set out [open frobots.txt w]
 		puts "creating robots.txt in $d"
 		close $out
+                incr status(unvisited)
 	    }
         }
     }
@@ -170,6 +177,9 @@ proc RobotFileOpen {area host path {mode w}} {
     } else {
         set out [open f $mode]
     }
+    if {$mode == "w"} {
+        incr status($area)
+    }
     cd $orgPwd
     return $out
 }
@@ -676,11 +686,6 @@ proc RobotWriteMetadata {url out} {
         text/plain {
             RobotTextPlain $url $out
         }
-        application/pdf {
-            set pdff [open test.pdf w]
-            puts -nonewline $pdff $URL($url,buf)
-            close $pdff
-        }
     }
     puts $out "</zmbot>"
 }
@@ -692,10 +697,6 @@ proc Robot200 {url} {
     puts -nonewline $out $URL($url,buf)
     RobotFileClose $out
 
-    if {![checkrule mime $URL($url,head,content-type)]} {
-        RobotError $url mimedeny
-        return
-    }
     set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
     RobotWriteMetadata $url $out
     RobotFileClose $out
@@ -730,6 +731,7 @@ proc RobotReadHeader {url sock} {
     if {[catch {set buffer [read $sock 2148]}]} {
 	RobotError $url 404
 	RobotRestart $url $sock
+        return
     }
     set readCount [string length $buffer]
     
@@ -770,12 +772,19 @@ proc RobotReadHeader {url sock} {
 		    if {![info exists URL($url,head,content-type)]} {
 			set URL($url,head,content-type) {}
 		    }
-		    set binary 0
-		    switch $URL($url,head,content-type) {
-			application/pdf {
-			    set binary 1
+		    set binary 1
+		    switch -glob -- $URL($url,head,content-type) {
+			text/* {
+			    set binary 0
 			}
 		    }
+                    if {![regexp {/robots.txt$} $url]} {
+                        if {![checkrule mime $URL($url,head,content-type)]} {
+                            RobotError $url mimedeny
+                            RobotRestart $url $sock
+                            return
+                        }
+                    }
 		    fileevent $sock readable [list RobotReadContent $url $sock $binary]
 		}
 		default {
@@ -892,6 +901,10 @@ set workdir [pwd]
 set idletime 60000
 set acceptLanguage {}
 set debuglevel 0
+set status(unvisited) 0
+set status(visited) 0
+set status(bad) 0
+set status(raw) 0
 
 
 # Rules: allow, deny, url
@@ -1061,9 +1074,12 @@ puts "domains=$domains"
 puts "max distance=$maxdistance"
 puts "max jobs=$robotsMax"
 
+
 RobotStart
 
 
 while {$robotsRunning} {
     vwait robotsRunning
 }
+
+puts "End un,ba,vi=$status(unvisited),$status(bad),$status(visited)"