index: escape colon characters inside ipaths. This could potentially happen with the zip (ie: zipped maildir) and chm filters

src/internfile/internfile.cpp

@@ -53,10 +53,31 @@ using namespace std;
 #include "pxattr.h"
 #endif // RCL_USE_XATTR
 
+static const string stxtplain("text/plain");
+
 // The internal path element separator. This can't be the same as the rcldb 
 // file to ipath separator : "|"
+// We replace it with a control char if it comes out of a filter (ie:
+// rclzip or rclchm can do this). If you want the SOH control char
+// inside an ipath, you're out of luck (and a bit weird).
 static const string isep(":");
-static const string stxtplain("text/plain");
+static const char colon_repl = '\x01';
+static string colon_hide(const string& in)
+{
+    string out;
+    for (string::const_iterator it = in.begin(); it != in.end(); it++) {
+	out += *it == ':' ? colon_repl : *it;
+    }
+    return out;
+}
+static string colon_restore(const string& in)
+{
+    string out;
+    for (string::const_iterator it = in.begin(); it != in.end(); it++) {
+	out += *it == colon_repl ? ':' : *it;
+    }
+    return out;
+}
 
 set<string> FileInterner::o_missingExternal;
 map<string, set<string> >  FileInterner::o_typesForMissing;
@@ -603,7 +624,7 @@ void FileInterner::collectIpathAndMT(Rcl::Doc& doc, string& ipath) const
 		getKeyValue(docdata, keymt, doc.mimetype);
 		getKeyValue(docdata, keyfn, doc.utf8fn);
 	    }
-	    ipath += ipathel + isep;
+	    ipath += colon_hide(ipathel) + isep;
 	} else {
 	    ipath += isep;
 	}
@@ -750,6 +771,10 @@ FileInterner::Status FileInterner::internfile(Rcl::Doc& doc, string& ipath)
     if (!ipath.empty()) {
 	vector<string> lipath;
 	stringToTokens(ipath, lipath, isep, true);
+	for (vector<string>::iterator it = lipath.begin();
+	     it != lipath.end(); it++) {
+	    *it = colon_restore(*it);
+	}
 	vipath.insert(vipath.begin(), lipath.begin(), lipath.end());
 	if (!m_handlers.back()->skip_to_document(vipath[m_handlers.size()-1])){
 	    LOGERR(("FileInterner::internfile: can't skip\n"));

src/internfile/mh_execm.cpp

@@ -167,6 +167,8 @@ bool MimeHandlerExecMultiple::next_document()
         obuf << "Filename: " << 0 << "\n";
     }
     if (m_ipath.length()) {
+	LOGDEB(("next_doc: sending len %d val [%s]\n", m_ipath.length(),
+		m_ipath.c_str()));
         obuf << "Ipath: " << m_ipath.length() << "\n" << m_ipath;
     }
     if (!m_dfltInputCharset.empty()) {

tests/mail/mail.sh

@@ -12,6 +12,7 @@ initvariables $0
   recollq '"Dear Corporate Administrator"'
   recollq TestTbirdWithoutEmptyLine
   recollq TestTbirdWithEmptyLine
+  recollq ZIPPEDMAILDIR_UNIQUEXXX
 ) 2> $mystderr | egrep -v '^Recoll query: ' > $mystdout
 
 diff -w ${myname}.txt $mystdout > $mydiffs 2>&1

tests/mail/mail.txt

@@ -11,3 +11,5 @@ message/rfc822	[file:///home/dockes/projets/fulltext/testrecoll/mail/badMail.edi
 message/rfc822	[file:///home/dockes/projets/fulltext/testrecoll/mail/thunderbird/Sent]	[Pronote: salut les genies.]	2565	bytes	
 1 results
 message/rfc822	[file:///home/dockes/projets/fulltext/testrecoll/mail/thunderbird/Sent]	[De mieux en mieux]	2565	bytes	
+1 results
+message/rfc822	[file:///home/dockes/projets/fulltext/testrecoll/mail/maildir.zip]	[[FreeBSD-Announce]  ZIPPEDMAILDIR_UNIQUEXXX 	FreeBSD Security Advisory FreeBSD-SA-04:17.procfs]	59171	bytes	

website/BUGS.html

@@ -37,6 +37,11 @@
 
       <h2><a name="b_latest">recoll 1.15</a></h2> 
         <ul>
+        <li>The rclzip filter can't handle archive members with a colon
+        (':') in the file name or path. The files are normally indexed and
+        can be searched for, but they can't be displayed (neither opened
+        nor previewed).</li>
+
         <li>After an upgrade, the recoll GUI sometimes crashes on
           startup. This is fixed by removing (back it up just in case)
           ~/.config/Recoll.org/recoll.conf, the QSettings storage for