cleaned up processing of metadata from diverse origins (doc,extattrs,localfields)

src/index/fsindexer.cpp

@@ -47,8 +47,10 @@
 #include "rclinit.h"
 #include "execmd.h"
 
-// When using extended attributes, we have to use the ctime. 
-// This is quite an expensive price to pay...
+
+// When using extended attributes, we have to use the ctime, because
+// this is all that gets set when the attributes are modified. This
+// is quite an expensive price to pay...
 #ifdef RCL_USE_XATTR
 #define RCL_STTIME st_ctime
 #else
@@ -420,22 +422,21 @@ void FsIndexer::localfieldsfromconf()
     vector<string> nmlst = attrs.getNames(cstr_null);
     for (vector<string>::const_iterator it = nmlst.begin();
          it != nmlst.end(); it++) {
-	attrs.get(*it, m_localfields[*it]);
+	string nm = m_config->fieldCanon(*it);
+	attrs.get(*it, m_localfields[nm]);
+	LOGDEB2(("FsIndexer::localfieldsfromconf: [%s]->[%s]\n",
+		 nm.c_str(), m_localfields[nm].c_str()));
     }
 }
 
-
-// 
 void FsIndexer::setlocalfields(const map<string, string>& fields, Rcl::Doc& doc)
 {
     for (map<string, string>::const_iterator it = fields.begin();
 	 it != fields.end(); it++) {
-        // Should local fields override those coming from the document
-        // ? I think not, but not too sure. We could also chose to
-        // concatenate the values ?
-        if (doc.meta.find(it->second) == doc.meta.end()) {
-            doc.meta[it->first] = it->second;
-	}
+        // Being chosen by the user, localfields override values from
+        // the filter. The key is already canonic (see
+        // localfieldsfromconf())
+	doc.meta[it->first] = it->second;
     }
 }
 
@@ -484,7 +485,9 @@ void FsIndexer::reapmetadata(const vector<MDReaper>& reapers, const string& fn,
 	}
 	string output;
 	if (ExecCmd::backtick(cmd, output)) {
-	    doc.meta[rp->fieldname] += string(" ") + output;
+	    // addmeta() creates or appends. fieldname is already
+	    // canonic (see above)
+	    doc.addmeta(rp->fieldname, output);
 	}
     }
 }
@@ -534,7 +537,7 @@ void *FsIndexerInternfileWorker(void * fsp)
     TempDir tmpdir;
     RclConfig myconf(*(fip->m_stableconfig));
 
-    InternfileTask *tsk;
+    InternfileTask *tsk = 0;
     for (;;) {
 	if (!tqp->take(&tsk)) {
 	    tqp->workerExit();

src/index/rclmonrcv.cpp

@@ -567,22 +567,22 @@ const char *RclIntf::event_name(int code)
     code &= ~(IN_ISDIR|IN_ONESHOT);
     switch (code) {
     case IN_ACCESS: return "IN_ACCESS";
-    case IN_MODIFY: return "IN_MODIFY";
     case IN_ATTRIB: return "IN_ATTRIB";
-    case IN_CLOSE_WRITE: return "IN_CLOSE_WRITE";
-    case IN_CLOSE_NOWRITE: return "IN_CLOSE_NOWRITE";
     case IN_CLOSE: return "IN_CLOSE";
-    case IN_OPEN: return "IN_OPEN";
-    case IN_MOVED_FROM: return "IN_MOVED_FROM";
-    case IN_MOVED_TO: return "IN_MOVED_TO";
-    case IN_MOVE: return "IN_MOVE";
+    case IN_CLOSE_NOWRITE: return "IN_CLOSE_NOWRITE";
+    case IN_CLOSE_WRITE: return "IN_CLOSE_WRITE";
     case IN_CREATE: return "IN_CREATE";
     case IN_DELETE: return "IN_DELETE";
     case IN_DELETE_SELF: return "IN_DELETE_SELF";
-    case IN_MOVE_SELF: return "IN_MOVE_SELF";
-    case IN_UNMOUNT: return "IN_UNMOUNT";
-    case IN_Q_OVERFLOW: return "IN_Q_OVERFLOW";
     case IN_IGNORED: return "IN_IGNORED";
+    case IN_MODIFY: return "IN_MODIFY";
+    case IN_MOVE: return "IN_MOVE";
+    case IN_MOVED_FROM: return "IN_MOVED_FROM";
+    case IN_MOVED_TO: return "IN_MOVED_TO";
+    case IN_MOVE_SELF: return "IN_MOVE_SELF";
+    case IN_OPEN: return "IN_OPEN";
+    case IN_Q_OVERFLOW: return "IN_Q_OVERFLOW";
+    case IN_UNMOUNT: return "IN_UNMOUNT";
     default: {
 	static char msg[50];
 	sprintf(msg, "Unknown event 0x%x", code);
@@ -599,6 +599,12 @@ bool RclIntf::addWatch(const string& path, bool)
     // CLOSE_WRITE is covered through MODIFY. CREATE is needed for mkdirs
     uint32_t mask = IN_MODIFY | IN_CREATE
         | IN_MOVED_FROM | IN_MOVED_TO | IN_DELETE
+#ifdef RCL_USE_XATTR
+	// It seems that IN_ATTRIB is not needed to receive extattr
+	// modification events, which is a bit weird because only ctime is
+	// set. 
+	// | IN_ATTRIB
+#endif
 #ifdef IN_DONT_FOLLOW
 	| IN_DONT_FOLLOW
 #endif
@@ -692,7 +698,7 @@ bool RclIntf::getEvent(RclMonEvent& ev, int msecs)
 	eraseWatchSubTree(m_idtopath, ev.m_path);
     }
 
-
+    // IN_ATTRIB apparently not needed, see comment above
     if (evp->mask & (IN_MODIFY)) {
 	ev.m_etyp = RclMonEvent::RCLEVT_MODIFY;
     } else if (evp->mask & (IN_DELETE | IN_MOVED_FROM)) {