Nabble - Opensync - Dev

New libsyncml tool: OBEX cutter

2008-07-04T04:50:18Z

Hi,

i wrote a tiny tool to cut the OBEX payload from captured SyncML traffic. For
easier debugging/reviewing traffic dumps - so you don't have to read OBEX
traffic to extract the wbxml2/SyncML bits.

obex-cutter could be used like this: Imagine you have a hexdump of a single
OBEX message:

libsyncml/tools> cat dump.hex
8200fdcb00000005c3000000cd4200206170706c69636174696f6e2f766e642e73796e636d6c2b7762786d6c004900d0029f536a0e73796e636d6c3a6d6574696e66006d6c7103312e310001720353796e634d4c2f312e31000165033100015b033100016e5703494d45493a46464646464646464646464646464600010167570350432053756974650001015a00014c03313034383537360001010100006b464b033100014f033230360001546e57032f74656c65636f6d2f70622e766366000101675703436f6e74616374730001015a00015303746578742f782d766361726400010100000f0101120101
libsyncml/tools> cat dump.hex | ./obex-cutter > dump.wbxml.hex

# In dump.wbxml.hex the OBEX payload got stripped. Only the wbxml2 compressed
SyncML message is left in hex. Convert it to binary:

libsyncml/tools> osyncbinary --hex dump.wbxml.hex dump.wbxml &> /dev/null

# Convert wbxml to XML

libsyncml/tools> wbxml2xml -o dump.xml dump.wbxml
wbxml2xml succeded

# Ready:

libsyncml/tools> head dump.xml
<?xml version="1.0"?>
<!DOCTYPE SyncML PUBLIC "-//SYNCML//DTD SyncML
1.1//EN" "http://www.syncml.org/docs/syncml_represent_v11_20020213.dtd">
<SyncML xmlns="syncml:SYNCML1.1">
<SyncHdr>
<VerDTD>1.1</VerDTD>
<VerProto>SyncML/1.1</VerProto>
<SessionID>1</SessionID>
<MsgID>1</MsgID>
<Target>
<LocURI>IMEI:FFFFFFFFFFFFFFF</LocURI>

Michael, good to go for libsyncml/tools?

best regards,
Daniel

[attachment removed]
-------------------------------------------------------------------------
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Svn kdepim does not compile

2008-06-30T03:57:52Z

make reports the following error:

/home/aeg/local/opensync/kdepim/src/datasource.cpp:162: error:
'osync_hashtable_report' was not declared in this scope

this is most likely because the according include line war removed
from opensync_hashtable.h

any hints how to fix this?

thanks

alexander
--
Dr. Alexander Egger, Studiengang IT & IT-Marketing
CAMPUS 02 Fachhochschule der Wirtschaft GmbH
Körblergasse 126, 8021 Graz, Austria/EUROPE
T:+43 316 6002 395 M:+43 699 14102026

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Re: osync_xmlformat_validate() in format plugins

2008-06-26T03:58:34Z

Hi Daniel,

You are welcome ;-)

> The idea of using C++-singleton-like sounds interesting. I wonder if we would
> change some more stuff, if this would be still required. For example make all
> OSyncXMLFormat "inherit" of OSyncXMLFormatSchema? We just force/require for
> all XMLFormats a schema file, by changing osync_xmlformat_new parameter list
> to:
>
> osync_xmlformat_new(const char *objtype, OSyncXMLFormatSchema *schema,
> OSyncError **error);
>
> And putting an OSyncXMLFormatSchema pointer into the internal struct.
>
> Thoughts?
>
>
>

The last days I was thinking about that and it is alway better not to

use any static variables because of concurrent access. But I don't think
we could avoid this problem without having more schema instances in the
memory. If we use your solution every format plugin has to create its
own instance of each schema which the plugin will support. This will be
better than today but not really nice.

Maybe you can review my code again and test it. I changed the xmlformat
test suite to fit with the new solution, too. If my code is ok I will
commit it. Afterwards I take a look at the format plugins and change
them if necessary.

>> /**
>> * @brief Validate the xmlformat against its schema in inidivual path
>> * @param xmlformat The pointer to a xmlformat object
>> @@ -399,8 +473,9 @@
>> osync_bool osync_xmlformat_validate(OSyncXMLFormat *xmlformat)
>> {
>> osync_assert(xmlformat);
>> -
>> - return _osync_xmlformat_validate(xmlformat, NULL);
>> +
>> + OSyncXMLFormatSchema * schema =
>> osync_xmlformat_schema_get_instance(xmlformat, NULL); + return
>> osync_xmlformat_schema_validate(schema, xmlformat);
>>
>
> We should avoid code which is ignoring the error handling by passing NULL
> instead of a OSyncError reference. (I know there is/was some code doing
> so ... but we should get rid of those old code. It's just bad/wrong/....).
> Feel free to change the paramter list vor the _validate() function to
> (OSyncXMLFormat *xmlformat, OSyncError **error). You might set also an error
> when osync_xmlformat_schema_validate is false:
> osync_error_set(*error, OSYNC_ERROR_GENERIC /*or something different */, "XML
> Format validation failed.");
>

changed.

--
/Bjoern Ricks

Index: opensync/xmlformat/opensync_xmlformat.c
===================================================================
--- opensync/xmlformat/opensync_xmlformat.c (Revision 3381)
+++ opensync/xmlformat/opensync_xmlformat.c (Arbeitskopie)
@@ -97,6 +97,7 @@
* @param path The individual schema path. If NULL the default OPENSYNC_SCHEMASDIR is used.
* @return TRUE if xmlformat valid else FALSE
*/
+/*
osync_bool _osync_xmlformat_validate(OSyncXMLFormat *xmlformat, const char *path)
{
osync_assert(xmlformat);
@@ -111,8 +112,141 @@
g_free(schemafilepath);

return res;
+} */
+
+/**
+ * @brief Create new OSyncXMLFormatSchema for xmlformat
+ * @param xmlformat The pointer to a xmlformat object. xmlformat->objtype is used to identify the schema file
+ * @param path The individual schema path. If NULL the default OPENSYNC_SCHEMASDIR is used.
+ * @param error The error which will hold the info in case of an error
+ * @return new OSyncXMLFormatSchema or NULL in case of an error
+ */
+OSyncXMLFormatSchema * osync_xmlformat_schema_new(OSyncXMLFormat * xmlformat, const char *path, OSyncError **error) {
+ OSyncXMLFormatSchema * osyncschema = NULL;
+
+ osync_trace(TRACE_ENTRY, "%s(%p, %p, %p)", __func__, xmlformat, path, error);
+
+ osync_assert(xmlformat);
+
+ osyncschema = osync_try_malloc0(sizeof(OSyncXMLFormatSchema), error);
+ if(!osyncschema) {
+ osync_trace(TRACE_EXIT_ERROR, "%s: %s" , __func__, osync_error_print(error));
+ // release mutex
+ return NULL;
+ }
+ osyncschema->objtype = g_strdup(osync_xmlformat_get_objtype(xmlformat));
+ char *schemafilepath = g_strdup_printf("%s%c%s%s%s",
+ path ? path : OPENSYNC_SCHEMASDIR,
+ G_DIR_SEPARATOR,
+ "xmlformat-",
+ osyncschema->objtype,
+ ".xsd");
+
+ xmlSchemaParserCtxtPtr xmlSchemaParserCtxt;
+ xmlSchemaPtr xmlSchema;
+ xmlSchemaValidCtxtPtr xmlSchemaValidCtxt;
+
+ xmlSchemaParserCtxt = xmlSchemaNewParserCtxt(schemafilepath);
+ osyncschema->schema = xmlSchemaParse(xmlSchemaParserCtxt);
+ xmlSchemaFreeParserCtxt(xmlSchemaParserCtxt);
+
+ osyncschema->context = xmlSchemaNewValidCtxt(osyncschema->schema);
+ if (osyncschema->context == NULL) {
+ xmlSchemaFree(osyncschema->schema);
+ g_free(osyncschema->objtype);
+ g_free(osyncschema);
+ osyncschema = NULL;
+ osync_error_set(error, OSYNC_ERROR_GENERIC, "XMLFormat validation failed. Could not create Schema Context.");
+ }
+ return osyncschema;
}

+/**
+ * @brief Free existing OSyncXMLFormatSchema
+ * @param osyncschema Pointer to the Schema that shoud be freed
+ * @param error The error which will hold the info in case of an error
+ */
+void osync_xmlformat_schema_free(OSyncXMLFormatSchema * osyncschema) {
+
+ osync_assert(osyncschema);
+
+ xmlSchemaFreeValidCtxt(osyncschema->context);
+ xmlSchemaFree(osyncschema->schema);
+ g_free(osyncschema->objtype);
+ g_free(osyncschema);
+
+}
+
+/**
+ * @brief Get a schema for the xmlformat.
+ *
+ * This function creates only one instance of a schema for each objtype. If a xmlformat is passed as a parameter with the same
+ * objtype as a xmlformat prior the returned pointer to a OSyncXMLFormatSchema instance is the same as before.
+ *
+ * @param xmlformat The pointer to a xmlformat object
+ * @param error The error which will hold the info in case of an error
+ * @return Pointer to a instance of OSyncXMLFormatSchema
+ */
+
+OSyncXMLFormatSchema * osync_xmlformat_schema_get_instance(OSyncXMLFormat * xmlformat, OSyncError **error) {
+ static GList * schemas = NULL;
+ static GStaticMutex mutex = G_STATIC_MUTEX_INIT;
+ OSyncXMLFormatSchema * osyncschema = NULL;
+ GList * entry;
+ const char * objtype;
+
+ osync_trace(TRACE_ENTRY, "%s(%p, %p)", __func__, xmlformat, error);
+
+ osync_assert(xmlformat);
+
+ objtype = osync_xmlformat_get_objtype(xmlformat);
+ // get mutex
+ g_static_mutex_lock(&mutex);
+ // find schema for objtype
+ for ( entry = schemas; entry != NULL; entry=entry->next) { // should be fast enough for only a few objtypes
+ osyncschema = (OSyncXMLFormatSchema *) entry->data;
+ if (!strcmp(osyncschema->objtype, objtype) ) {
+ return osyncschema;
+ }
+ osyncschema = NULL;
+ }
+ if ( osyncschema == NULL ) {
+ osyncschema = osync_xmlformat_schema_new(xmlformat, NULL, error);
+ if ( osyncschema != NULL ) {
+ schemas = g_list_append(schemas, osyncschema);
+ }
+ }
+ // release mutex
+ g_static_mutex_unlock(&mutex);
+ return osyncschema;
+}
+
+/**
+ * @brief Validate the xmlformat against its schema
+ * @param xmlformat The pointer to a xmlformat object
+ * @param schema The pointer to a OSyncXMLFormatSchema object.
+ * @param error The error which will hold the info in case of an error
+ * @return TRUE if xmlformat valid else FALSE
+ */
+
+osync_bool osync_xmlformat_schema_validate(OSyncXMLFormatSchema * schema, OSyncXMLFormat * xmlformat, OSyncError **error)
+{
+ osync_assert(xmlformat);
+ osync_assert(schema);
+
+ int rc = 0;
+
+ /* Validate the document */
+ rc = xmlSchemaValidateDoc(schema->context, xmlformat->doc);
+
+ if(rc != 0) {
+ osync_error_set(error, OSYNC_ERROR_GENERIC, "XMLFormat validation failed.");
+ return FALSE;
+ }
+ return TRUE;
+}
+
+
/*@}*/

/**
@@ -394,13 +528,15 @@
/**
* @brief Validate the xmlformat against its schema
* @param xmlformat The pointer to a xmlformat object
+ * @param error The error which will hold the info in case of an error
* @return TRUE if xmlformat valid else FALSE
*/
-osync_bool osync_xmlformat_validate(OSyncXMLFormat *xmlformat)
+osync_bool osync_xmlformat_validate(OSyncXMLFormat *xmlformat, OSyncError **error)
{
osync_assert(xmlformat);
-
- return _osync_xmlformat_validate(xmlformat, NULL);
+
+ OSyncXMLFormatSchema * schema = osync_xmlformat_schema_get_instance(xmlformat, error);
+ return osync_xmlformat_schema_validate(schema, xmlformat, error);
}

/**
Index: opensync/xmlformat/opensync_xmlformat_internals.h
===================================================================
--- opensync/xmlformat/opensync_xmlformat_internals.h (Revision 3381)
+++ opensync/xmlformat/opensync_xmlformat_internals.h (Arbeitskopie)
@@ -23,7 +23,9 @@
#ifndef OPENSYNC_XMLFORMAT_INTERNALS_H_
#define OPENSYNC_XMLFORMAT_INTERNALS_H_

-#include <opensync/opensync_xml.h>
+#include <libxml/xpath.h>
+#include <libxml/xmlschemas.h>
+#include <libxml/tree.h>

/**
* @brief Represent a XMLFormat object
@@ -45,7 +47,24 @@

};

+/**
+ * @brief Represents a Schema object
+ * @ingroup OSyncXMLFormatPrivateAPI
+ */
+typedef struct OSyncXMLFormatSchema {
+ /** The schema object */
+ xmlSchemaPtr schema;
+ /** The schema validation context */
+ xmlSchemaValidCtxtPtr context;
+ /** The object type of OSyncXMLFormat */
+ char * objtype;
+} OSyncXMLFormatSchema;
+
int _osync_xmlformat_get_points(OSyncXMLPoints points[], int* cur_pos, int basic_points, const char* fieldname);
-osync_bool _osync_xmlformat_validate(OSyncXMLFormat *xmlformat, const char *path);

+OSyncXMLFormatSchema * osync_xmlformat_schema_get_instance(OSyncXMLFormat * xmlformat, OSyncError **error);
+OSyncXMLFormatSchema * osync_xmlformat_schema_new(OSyncXMLFormat * xmlformat, const char *path, OSyncError **error);
+void osync_xmlformat_schema_free(OSyncXMLFormatSchema * schema);
+osync_bool osync_xmlformat_schema_validate(OSyncXMLFormatSchema * schema, OSyncXMLFormat * xmlformat, OSyncError **error);
+
#endif /*OPENSYNC_XMLFORMAT_INTERNAL_H_*/
Index: opensync/xmlformat/opensync_xmlformat.h
===================================================================
--- opensync/xmlformat/opensync_xmlformat.h (Revision 3381)
+++ opensync/xmlformat/opensync_xmlformat.h (Arbeitskopie)
@@ -48,7 +48,7 @@
OSYNC_EXPORT OSyncXMLFieldList *osync_xmlformat_search_field(OSyncXMLFormat *xmlformat, const char *name, OSyncError **error, ...);

OSYNC_EXPORT osync_bool osync_xmlformat_assemble(OSyncXMLFormat *xmlformat, char **buffer, unsigned int *size);
-OSYNC_EXPORT osync_bool osync_xmlformat_validate(OSyncXMLFormat *xmlformat);
+OSYNC_EXPORT osync_bool osync_xmlformat_validate(OSyncXMLFormat *xmlformat, OSyncError **error);

OSYNC_EXPORT void osync_xmlformat_sort(OSyncXMLFormat *xmlformat);

Index: tests/merger-tests/check_xmlformat.c
===================================================================
--- tests/merger-tests/check_xmlformat.c (Revision 3381)
+++ tests/merger-tests/check_xmlformat.c (Arbeitskopie)
@@ -275,6 +275,7 @@
char *buffer;
unsigned int size;
OSyncError *error = NULL;
+ OSyncXMLFormatSchema *schema = NULL;

fail_unless(osync_file_read("event.xml", &buffer, &size, &error), NULL);
fail_unless(error == NULL, NULL);
@@ -283,9 +284,13 @@
fail_unless(error == NULL, NULL);

g_free(buffer);
-
- fail_unless(_osync_xmlformat_validate(xmlformat, testbed) != FALSE, NULL);
-
+ schema = osync_xmlformat_schema_new(xmlformat, testbed, &error);
+ fail_if( schema == NULL );
+ //fail_unless(_osync_xmlformat_validate(xmlformat, testbed) != FALSE, NULL);
+ fail_unless( osync_xmlformat_schema_validate(schema, xmlformat, &error) );
+
+ osync_xmlformat_schema_free(schema);
+
osync_xmlformat_unref(xmlformat);

destroy_testbed(testbed);

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Re: libsyncml HTTP client support?

2008-06-25T09:45:55Z

Michael Bell wrote:
> Hi Josh,
>
> I did not make a real-life test with your code but I understand what's
> your problem :)
I had tried this change during testing, assuming what you explained
below to be the case. It still doesn't work right but I'm not sure why
exactly. The flow of events looks like:

SML_MANAGER_SESSION_NEW
SML_MANAGER_SESSION_FLUSH
SML_MANAGER_CONNECT_DONE
ds_client_init_sync_mode()
ds_client_get_changeinfo()
SML_MANAGER_SESSION_FINAL
smlDsSessionDispatch(0x8062950)
SML_MANAGER_SESSION_FLUSH
_recv_alert_reply(status=200)

Using a breakpoint, I find that smlDsSessionRecvAlert is not called
until some point after _recv_alert_reply, however no dispatch attempts
are made after the one on/after SML_MANAGER_SESSION_FINAL. No idea how
to proceed or why my event loop is not processing correctly.

>-----> snip >----->
> P.S. yes I know that I have to cleanup the SyncML DS client code which
> still makes useless calls to SetConnectCallback :(
Oh, OK. I had thought that must be true, but it's nice to have it
confirmed so I can stop investigating that section of the code :)

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Re: libsyncml HTTP client support?

2008-06-25T01:19:52Z

Hi Josh,

I did not make a real-life test with your code but I understand what's
your problem :)

Josh Guilfoyle wrote:

> I understand this piece as you can see more clearly in my code. My
> point, however, is that without the SetConnectCallback callback firing,
> I have no opportunity to fit the specific SmlDsSession into my event
> loop. That is what I meant by RecvAlert being called, but the event is
> never dispatched as smlDsSessionDispatch is not being called correctly
> by my event loop. I am not 100% sure that this is the only problem at
> play, though. Looking through the code I am very confused by how
> SetConnectCallback and smlDsServerSendAlert work together. SendAlert
> seems to override the callback that SetConnectCallback had installed
> with the manager, but the SyncML plugin doesn't seem to care about this
> somehow.

First Ds means DatastoreServer and not DatasynchronizationServer. I know
that this is really dopey but I don't want to break the whole API to fix
this.

Second some explanations on the API logic - the interesting part ;)

If you create the DsServer then you must have a chance to get access to
a remotely initiated DsSession (especially if your an OMA DS server).
This is the use case of SetConnectCallback. The callback has nothing to
do with the transport connect. If you are an OMA DS client then you
don't need this callback.

If you are the OMA DS client then you get the DsSession from SendAlert.
Therefore this function replaces DsServerRecvAlert by DsSessionRecvAlert
at the manager (because there is already a DsSession).

So SendAlert sets firstds in your code and this is the place where you
get the used DsSession. You can start calling smlDsSessionDispatch
beginning with this function call.

So what does this means?

You added the required code by calling _ds_alert which does nothing else
then adding firstds to the array sessions. Your workaround is the
correct API use ;)

I hope this helps you a little bit.

Best regards

Michael

P.S. yes I know that I have to cleanup the SyncML DS client code which
still makes useless calls to SetConnectCallback :(
--
_______________________________________________________________

Michael Bell Humboldt-Universitaet zu Berlin

Tel.: +49 (0)30-2093 2482 ZE Computer- und Medienservice
Fax: +49 (0)30-2093 2704 Unter den Linden 6
michael.bell@... D-10099 Berlin
_______________________________________________________________

X.509 CA Certificates / Wurzelzertifikate

https://pki.pca.dfn.de/hu-berlin-ca/cgi-bin/pub/pki?cmd=getStaticPage;name=index;id=2&RA_ID=0

--- syncml-client.c.org 2008-06-25 10:03:59.999703658 +0200
+++ syncml-client.c 2008-06-25 10:18:04.395754210 +0200
@@ -133,7 +133,7 @@
last, next, _recv_alert_reply, NULL, &err);
assert(firstds != NULL);

- _ds_alert(firstds, NULL);
+ sessions = g_list_append(sessions, firstds);

smlDsSessionGetAlert(firstds, _ds_client_recv_alert, NULL);
smlSessionFlush(first, TRUE, NULL);

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

smime.p7s (7K) Download Attachment

Re: libsyncml HTTP client support?

2008-06-23T13:13:04Z

Yes, sorry, the attachment was missing. I've corrected that here. See
below for further response.

Michael Bell wrote:

> Josh Guilfoyle wrote:
>> I've been looking into this more and ran into a problem that is not
>> apparent from the opensync plugin. I have attached code I extracted
>> in order to form a very simple functional test of libsyncml operating
>> as a client to my server, also using libsyncml. You should be able
>> to trivially modify it to connect to your own local test server and I
>> expect you will see the same results.
>
> I think the attachment is missing ;)
>
> Before I start a short warning: I do not develop the code. I only
> interpret it. So I'm sorry if I'm not fully correct and I hope some
> more longstanding developers of OpenSync read and check this reply too.
>

Thanks, I really appreciate the effort to help me here. There probably

aren't a lot of folks using libsyncml as I am, so I can appreciate that
my questions seem odd.

>> What is happening is that the client is connecting, sending the
>> <Alert>, which is responded to by the server, which also sends its
>> anticipated <Alert>. However, _ds_client_recv_alert is not
>> triggered. Analyzing the log file, I see that smlDsSessionRecvAlert
>> *is* called, but no corresponding smlDsSessionDispatch to get the
>> callback and response processed. Similarly, _ds_alert (from
>> smlDsServerSetConnectCallback) is never triggered, which could be the
>> reason why smlDsSessionDispatch didn't occur as expected.
>
> smlDsSessionRecvAlert is only used to signal the received alert. This
> function is usually registered at the manager and called by the
> manager. This function mainly build the response.
>
> smlDsSessionDispatch is the real dispatcher. libsyncml uses glib's
> mechanisms for event management here. It creates a GSource and
> registers it together with some functions at a context. glib loops on
> the context and calls the dispatch function from time to time.
>
> If you don't want to implement it this way then you can write some
> kind of while or for loop which call smlDsSessionDispatch.

I understand this piece as you can see more clearly in my code. My
point, however, is that without the SetConnectCallback callback firing,
I have no opportunity to fit the specific SmlDsSession into my event
loop. That is what I meant by RecvAlert being called, but the event is
never dispatched as smlDsSessionDispatch is not being called correctly
by my event loop. I am not 100% sure that this is the only problem at
play, though. Looking through the code I am very confused by how
SetConnectCallback and smlDsServerSendAlert work together. SendAlert
seems to override the callback that SetConnectCallback had installed
with the manager, but the SyncML plugin doesn't seem to care about this
somehow.

Anyway, the code should make it more clear how I am utilizing this area
of libsyncml. I tried to copy the opensync syncml plugin closely to
build a simple test.
>
> @Daniel: I hope the explanation is correct. I'm not a glib expert. I
> only read and interpret the syncml plugin code.
>
> Best regards
>
> Michael

/*
* $Id$
*
* Copyright (C) 2008 Josh Guilfoyle <jasta@...>
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2, or (at your option) any
* later version.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*/

#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include <assert.h>

#define __STDC_FORMAT_MACROS
#include <inttypes.h> /* PRIx32 */

#include <glib.h>
#include <glib-object.h>

#include <libsyncml/syncml.h>
#include <libsyncml/sml_auth.h>
#include <libsyncml/sml_ds_server.h>

#include <libsyncml/http_client.h>

#include <libsyncml/sml_devinf_obj.h>

/*****************************************************************************/

#define DBNAME "music"
#define DBTYPE "application/x-fivedb"
#define SERVERURL "http://10.1.0.175:5545/sync"

/* Hehe, lame. Just for this test... */
static SmlTransport *tsp = NULL;
static SmlManager *mgr = NULL;
static SmlDevInfAgent *agent = NULL;
static SmlDsServer *dscli = NULL;
static SmlSession *first = NULL;
static SmlDsSession *firstds = NULL;

static GList *sessions = NULL;

/*****************************************************************************/

static void ds_client_get_changeinfo(void);
static SmlBool _ds_client_recv_alert(SmlDsSession *dsess, SmlAlertType type,
const char *last, const char *next, gpointer udata);
static void _ds_alert(SmlDsSession *dsess, gpointer udata);

static void _recv_alert_reply(SmlSession *sess, SmlStatus *status, void *udata)
{
unsigned int code = smlStatusGetCode(status);
printf("_recv_alert_reply(status=%u)\n", code);
smlDsSessionDispatch(firstds);
}

static SmlBool _ds_client_recv_alert(SmlDsSession *dsess, SmlAlertType type,
const char *last, const char *next, gpointer udata)
{
printf("I BELIEVE WE NEED TO DO SOMETHING, BUT WHAT!?\n");
return TRUE;
}

static void _recv_sync_reply(SmlSession *sess, SmlStatus *status, void *udata)
{
unsigned int code = smlStatusGetCode(status);
printf("_recv_sync_reply(status=%u)\n", code);
}

static void _ds_event(SmlDsSession *dsess, SmlDsEvent ev, void *udata)
{
printf("_ds_event: ?\n");
}

static void _recv_sync(SmlDsSession *dsess, unsigned int nchanges, void *udata)
{
printf("_recv_sync(nchanges=%u)\n", nchanges);
}

static SmlBool _recv_change(SmlDsSession *dsess, SmlChangeType type,
const char *uid, char *data, unsigned int size, const char *contenttype,
void *udata, SmlError **err)
{
printf("_recv_change(type=%d, uid=%s, mime=%s)\n", (int)type,
uid, contenttype);

return TRUE;
}

static void ds_client_get_changeinfo(void)
{
SmlError *err = NULL;

printf("ds_client_get_changeinfo()\n");

smlDsSessionGetEvent(firstds, _ds_event, NULL);
smlDsSessionGetSync(firstds, _recv_sync, NULL);
smlDsSessionGetChanges(firstds, _recv_change, NULL);

if (smlDsSessionSendSync(firstds, 0, _recv_sync_reply, NULL, &err) == FALSE)
abort();

if (smlDsSessionCloseSync(firstds, &err) == FALSE)
abort();

smlSessionFlush(first, TRUE, NULL);
}

static void ds_client_init_sync_mode(void)
{
SmlError *err = NULL;
char *last = NULL;
char next[17];
time_t htime;

printf("ds_client_init_sync_mode()\n");

htime = time(NULL);
strftime(next, sizeof(next), "%s", gmtime(&htime));

assert(firstds == NULL);
firstds = smlDsServerSendAlert(dscli, first, SML_ALERT_ONE_WAY_FROM_SERVER,
last, next, _recv_alert_reply, NULL, &err);
assert(firstds != NULL);

_ds_alert(firstds, NULL);

smlDsSessionGetAlert(firstds, _ds_client_recv_alert, NULL);
smlSessionFlush(first, TRUE, NULL);

ds_client_get_changeinfo();
}

static void _manager_event(SmlManager *mgr, SmlManagerEventType type,
SmlSession *sess, SmlError *assocerr, gpointer udata)
{
SmlError *err = NULL;

g_warning("_manager_event: TODO");

switch (type)
{
case SML_MANAGER_SESSION_FLUSH:
printf("SML_MANAGER_SESSION_FLUSH\n");
break;

/* To get here, we just set up the basic connection mechanics, now we
* must initiate the SyncML session and start the event queue engine
* moving. */
case SML_MANAGER_CONNECT_DONE:
printf("SML_MANAGER_CONNECT_DONE\n");
ds_client_init_sync_mode();
break;

case SML_MANAGER_DISCONNECT_DONE:
printf("SML_MANAGER_DISCONNECT_DONE\n");
break;

case SML_MANAGER_TRANSPORT_ERROR:
printf("SML_MANAGER_TRANSPORT_ERROR: %s\n",
smlErrorPrint(&assocerr));
smlTransportDisconnect(tsp, NULL, NULL);
break;

case SML_MANAGER_SESSION_NEW:
printf("SML_MANAGER_SESSION_NEW\n");
smlSessionUseStringTable(sess, TRUE);
smlSessionUseNumberOfChanges(sess, TRUE);
smlSessionSetReceivingMaxMsgSize(sess, 4096);
smlSessionUseLargeObjects(sess, TRUE);
smlSessionRef(sess);
break;

case SML_MANAGER_SESSION_FINAL:
printf("SML_MANAGER_SESSION_FINAL\n");
if (smlSessionFlush(sess, TRUE, &err) == FALSE)
{
g_warning("smlSessionFlush failed: %s\n",
smlErrorPrint(&err));
smlErrorDeref(&err);
}
break;

case SML_MANAGER_SESSION_END:
printf("SML_MANAGER_SESSION_END\n");
smlTransportDisconnect(tsp, NULL, NULL);
break;

case SML_MANAGER_SESSION_ERROR:
printf("SML_MANAGER_SESSION_ERROR: %s, %s\n",
sess ? smlSessionGetSessionID(sess) : "(null)",
smlErrorPrint(&assocerr));
break;

case SML_MANAGER_SESSION_WARNING:
printf("SML_MANAGER_SESSION_WARNING: %s\n",
smlErrorPrint(&assocerr));
break;

default:
abort();
break;
}
}

static void _ds_alert(SmlDsSession *dsess, gpointer udata)
{
g_message("Got here, I think we did the right thing...\n");

smlDsSessionRef(dsess);

sessions = g_list_append(sessions, dsess);
}

static gboolean client_init(void)
{
SmlDevInf *devinf;
SmlLocation *loc = NULL;
SmlTransportHttpClientConfig cfg;
SmlAuthenticator *auth;
SmlError *err = NULL;

tsp = smlTransportNew(SML_TRANSPORT_HTTP_CLIENT, &err);
assert(tsp != NULL);

mgr = smlManagerNew(tsp, &err);
assert(mgr != NULL);

smlManagerSetEventCallback(mgr, _manager_event, NULL);

/* AUTH */
auth = smlAuthNew(&err);
assert(auth != NULL);

smlAuthSetEnable(auth, FALSE);

if (smlAuthRegister(auth, mgr, &err) == FALSE)
abort();

/* DEVINF */
devinf = smlDevInfNew("syncml-client", SML_DEVINF_DEVTYPE_WORKSTATION, &err);
assert(devinf != NULL);

smlDevInfSetSoftwareVersion(devinf, "0.0.1");
smlDevInfSetSupportsNumberOfChanges(devinf, TRUE);
smlDevInfSetSupportsLargeObjs(devinf, TRUE);
smlDevInfSetSupportsUTC(devinf, TRUE);

agent = smlDevInfAgentNew(devinf, &err);
assert(agent != NULL);

if (smlDevInfAgentRegister(agent, mgr, &err) == FALSE)
return FALSE;

/* DATABASE */
loc = smlLocationNew(DBNAME, NULL, &err);
assert(loc != NULL);

dscli = smlDsClientNew(DBTYPE, loc, loc, &err);
assert(dscli != NULL);

if (smlDsServerRegister(dscli, mgr, &err) == FALSE)
abort();

smlDsServerSetConnectCallback(dscli, _ds_alert, NULL);

/* GO! */
if (smlManagerStart(mgr, &err) == FALSE)
abort();

memset(&cfg, 0, sizeof(cfg));
cfg.url = SERVERURL;

if (smlTransportInitialize(tsp, &cfg, &err) == FALSE)
abort();

return TRUE;
}

static gboolean _sessions_prepare(GSource *src, gint *timeout)
{
*timeout = 50;
return FALSE;
}

static gboolean _sessions_check(GSource *src)
{
GList *s;

for (s = sessions; s != NULL; s = s->next)
{
if (smlDsSessionCheck(s->data) == TRUE)
return TRUE;
}

if (smlManagerCheck(mgr) == TRUE)
return TRUE;

return FALSE;
}

static gboolean _sessions_dispatch(GSource *src, GSourceFunc cb, gpointer udata)
{
GList *s;

printf("_sessions_dispatch()\n");

for (s = sessions; s != NULL; s = s->next)
{
printf("smlDsSessionDispathch(%p)\n", s->data);
smlDsSessionDispatch(s->data);
}

smlManagerDispatch(mgr);

return TRUE;
}

static void client_attach(void)
{
GSourceFuncs *funcs = g_malloc0(sizeof(GSourceFuncs));
funcs->prepare = _sessions_prepare;
funcs->check = _sessions_check;
funcs->dispatch = _sessions_dispatch;
funcs->finalize = NULL;

GSource *src = g_source_new(funcs, sizeof(GSource));
g_source_set_callback(src, NULL, NULL, NULL);
g_source_attach(src, NULL);
}

char *generate_session_id(void)
{
char *buf;
guint32 r;

buf = malloc(17);
assert(buf != NULL);

/* Pack the randomness into a buffer for encoding. */
r = g_random_int();
sprintf(buf, "%08" PRIx32, (uint32_t)r);

r = g_random_int();
sprintf(buf + 8, "%08" PRIx32, (uint32_t)r);

return buf;
}

static char *client_connect(void)
{
SmlError *err = NULL;
char *sessId;
SmlLocation *target;
SmlLocation *source;
SmlLink *link;

sessId = generate_session_id();
printf("Initializing session %s\n", sessId);

target = smlLocationNew(DBNAME, NULL, &err);
assert(target != NULL);

source = smlLocationNew(DBNAME, NULL, &err);
assert(source != NULL);

link = smlLinkNew(tsp, NULL, &err);
assert(link != NULL);

assert(first == NULL);
first = smlSessionNew(SML_SESSION_TYPE_CLIENT, SML_MIMETYPE_WBXML,
SML_VERSION_11, SML_PROTOCOL_SYNCML, target, source, sessId, 0, &err);
assert(first != NULL);

/* We'll need to manage this ref. */
smlSessionRef(first);

if (smlManagerSessionAdd(mgr, first, link, &err) == FALSE)
abort();

if (smlDevInfAgentRegisterSession(agent, mgr, first, &err) == FALSE)
abort();

if (smlDevInfAgentSendDevInf(agent, first, &err) == FALSE)
abort();

if (smlDevInfAgentRequestDevInf(agent, first, &err) == FALSE)
abort();

if (smlSessionFlush(first, TRUE, &err) == FALSE)
abort();

return sessId;
}

static void client_finish(void)
{
smlSessionUnref(first);
}

/*****************************************************************************/

int main(int argc, char **argv)
{
GMainLoop *loop;
char *sessId;

g_type_init();

client_init();
client_attach();

sessId = client_connect();
printf("sessId=%s\n", sessId);

g_free(sessId);

g_message("Let's roll!");

loop = g_main_loop_new(NULL, FALSE);
assert(loop != NULL);

g_main_loop_run(loop);

client_finish();

return EXIT_SUCCESS;
}

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Re: osync_xmlformat_validate() in format plugins

2008-06-23T11:10:56Z

> On Thursday 19 June 2008 16:18:02 Bjoern Ricks wrote:
>
>> I started coding to get rid of osync_xml_validate and to have only one
>> instance of a xml schema for each objtype. Therefore I tried to reuse
>> the oop singleton pattern in C via a static GList. Could you please
>> review my first intuition because I am more familiar with C++ than C.
>>
>
> Thanks for looking into this!
>
> The idea of using C++-singleton-like sounds interesting. I wonder if we would
> change some more stuff, if this would be still required. For example make all
> OSyncXMLFormat "inherit" of OSyncXMLFormatSchema? We just force/require for
> all XMLFormats a schema file, by changing osync_xmlformat_new parameter list
> to:
>
> osync_xmlformat_new(const char *objtype, OSyncXMLFormatSchema *schema,
> OSyncError **error);
>
> And putting an OSyncXMLFormatSchema pointer into the internal struct.
>
> Thoughts?
>
> [...]
>

That could be a better solution than to use static variables for each
objtype. I was thinking about that too. But I didn't dare to change the
parameter list.
But the new solution has another problem. Where should the
OSyncXMLFormatSchemas be created (and released)? We need only one
instance per objtype. Is there one osnyc engine per application? Then
the engine could be the right part to create OSyncXMLFormatSchemas.

>
>
>> /**
>> * @brief Validate the xmlformat against its schema in inidivual path
>> * @param xmlformat The pointer to a xmlformat object
>> @@ -399,8 +473,9 @@
>> osync_bool osync_xmlformat_validate(OSyncXMLFormat *xmlformat)
>> {
>> osync_assert(xmlformat);
>> -
>> - return _osync_xmlformat_validate(xmlformat, NULL);
>> +
>> + OSyncXMLFormatSchema * schema =
>> osync_xmlformat_schema_get_instance(xmlformat, NULL); + return
>> osync_xmlformat_schema_validate(schema, xmlformat);
>>
>
> We should avoid code which is ignoring the error handling by passing NULL
> instead of a OSyncError reference. (I know there is/was some code doing
> so ... but we should get rid of those old code. It's just bad/wrong/....).
> Feel free to change the paramter list vor the _validate() function to
> (OSyncXMLFormat *xmlformat, OSyncError **error). You might set also an error
> when osync_xmlformat_schema_validate is false:
> osync_error_set(*error, OSYNC_ERROR_GENERIC /*or something different */, "XML
> Format validation failed.");
>

:-) I was irritated because osync_xmlformat_validate had no error
parameter. Therefore I had to use a NULL value. So we should change the
function signature.
--
------------------------------------------------------------------------
University of Applied Sciences Osnabrueck
Faculty of Engineering and Computer Science
Dipl.-Inf. (FH) Bjoern Ricks

P.O. Box 1940 - 49009 Osnabrueck - Germany
Phone: +49 541/969-3452 Fax: +49 541/969-13452
Email: b.ricks@... <mailto:b.ricks@...>
Web: http://www.ecs.fh-osnabrueck.de/ricks.html

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Re: osync_xmlformat_validate() in format plugins

2008-06-23T07:25:22Z

On Thursday 19 June 2008 16:18:02 Bjoern Ricks wrote:
> I started coding to get rid of osync_xml_validate and to have only one
> instance of a xml schema for each objtype. Therefore I tried to reuse
> the oop singleton pattern in C via a static GList. Could you please
> review my first intuition because I am more familiar with C++ than C.

Thanks for looking into this!

The idea of using C++-singleton-like sounds interesting. I wonder if we would
change some more stuff, if this would be still required. For example make all
OSyncXMLFormat "inherit" of OSyncXMLFormatSchema? We just force/require for
all XMLFormats a schema file, by changing osync_xmlformat_new parameter list
to:

osync_xmlformat_new(const char *objtype, OSyncXMLFormatSchema *schema,
OSyncError **error);

And putting an OSyncXMLFormatSchema pointer into the internal struct.

Thoughts?

[...]

> /**
> * @brief Validate the xmlformat against its schema in inidivual path
> * @param xmlformat The pointer to a xmlformat object
> @@ -399,8 +473,9 @@
> osync_bool osync_xmlformat_validate(OSyncXMLFormat *xmlformat)
> {
> osync_assert(xmlformat);
> -
> - return _osync_xmlformat_validate(xmlformat, NULL);
> +
> + OSyncXMLFormatSchema * schema =
> osync_xmlformat_schema_get_instance(xmlformat, NULL); + return
> osync_xmlformat_schema_validate(schema, xmlformat);

We should avoid code which is ignoring the error handling by passing NULL
instead of a OSyncError reference. (I know there is/was some code doing
so ... but we should get rid of those old code. It's just bad/wrong/....).
Feel free to change the paramter list vor the _validate() function to
(OSyncXMLFormat *xmlformat, OSyncError **error). You might set also an error
when osync_xmlformat_schema_validate is false:
osync_error_set(*error, OSYNC_ERROR_GENERIC /*or something different */, "XML
Format validation failed.");

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Re: libsyncml HTTP client support?

2008-06-23T06:26:45Z

Josh Guilfoyle wrote:
> I've been looking into this more and ran into a problem that is not
> apparent from the opensync plugin. I have attached code I extracted in
> order to form a very simple functional test of libsyncml operating as a
> client to my server, also using libsyncml. You should be able to
> trivially modify it to connect to your own local test server and I
> expect you will see the same results.

I think the attachment is missing ;)

Before I start a short warning: I do not develop the code. I only
interpret it. So I'm sorry if I'm not fully correct and I hope some more
longstanding developers of OpenSync read and check this reply too.

> What is happening is that the client is connecting, sending the <Alert>,
> which is responded to by the server, which also sends its anticipated
> <Alert>. However, _ds_client_recv_alert is not triggered. Analyzing
> the log file, I see that smlDsSessionRecvAlert *is* called, but no
> corresponding smlDsSessionDispatch to get the callback and response
> processed. Similarly, _ds_alert (from smlDsServerSetConnectCallback) is
> never triggered, which could be the reason why smlDsSessionDispatch
> didn't occur as expected.

smlDsSessionRecvAlert is only used to signal the received alert. This
function is usually registered at the manager and called by the manager.
This function mainly build the response.

smlDsSessionDispatch is the real dispatcher. libsyncml uses glib's
mechanisms for event management here. It creates a GSource and registers
it together with some functions at a context. glib loops on the context
and calls the dispatch function from time to time.

If you don't want to implement it this way then you can write some kind
of while or for loop which call smlDsSessionDispatch.

@Daniel: I hope the explanation is correct. I'm not a glib expert. I
only read and interpret the syncml plugin code.

Best regards

Michael
--
_______________________________________________________________

Michael Bell Humboldt-Universitaet zu Berlin

Tel.: +49 (0)30-2093 2482 ZE Computer- und Medienservice
Fax: +49 (0)30-2093 2704 Unter den Linden 6
michael.bell@... D-10099 Berlin
_______________________________________________________________

X.509 CA Certificates / Wurzelzertifikate

https://pki.pca.dfn.de/hu-berlin-ca/cgi-bin/pub/pki?cmd=getStaticPage;name=index;id=2&RA_ID=0

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

smime.p7s (7K) Download Attachment

Re: Plugin Building

2008-06-22T23:44:52Z

Hi Alexander,

we introduced a cmake makro that forbids to build within the source
dirs. Please create a separate directory like "build" and use it
instead. With a separate build dir it is easier to remove all files that
a generated by cmake (like the Makefiles or the CMakeFiles sub dirs). It
is possible to use only rm -r * in your build dir to clean up all cmake
files. Without a separate build dir you have to remove all cmake files
within the source directories.

> Hi,
>
> when running cmake on the kde-pim or google-calendar plugins from svn cmake
> throws an error, telling me that the build dir has to be different from the
> src dir.
>
> Is there still a reason for that or is that just a legacy?
>
> It would be nice to get rid of that because without that I can create
> eclipse CDT projects easier with cmake.
>
> Alexander
>
>

--
/Bjoern Ricks

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Plugin Building

2008-06-22T05:32:59Z

Hi,

when running cmake on the kde-pim or google-calendar plugins from svn cmake
throws an error, telling me that the build dir has to be different from the
src dir.

Is there still a reason for that or is that just a legacy?

It would be nice to get rid of that because without that I can create
eclipse CDT projects easier with cmake.

Alexander

--
Dr. Alexander Egger, Studiengang IT & IT-Marketing
CAMPUS 02 Fachhochschule der Wirtschaft GmbH
Körblergasse 126, 8021 Graz, Austria/EUROPE
T:+43 316 6002 395 M:+43 699 14102026

----------------------
GTDInbox.com: [P/Privat/opensync]

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Re: libsyncml HTTP client support?

2008-06-20T15:17:34Z

I've been looking into this more and ran into a problem that is not
apparent from the opensync plugin. I have attached code I extracted in
order to form a very simple functional test of libsyncml operating as a
client to my server, also using libsyncml. You should be able to
trivially modify it to connect to your own local test server and I
expect you will see the same results.

What is happening is that the client is connecting, sending the <Alert>,
which is responded to by the server, which also sends its anticipated
<Alert>. However, _ds_client_recv_alert is not triggered. Analyzing
the log file, I see that smlDsSessionRecvAlert *is* called, but no
corresponding smlDsSessionDispatch to get the callback and response
processed. Similarly, _ds_alert (from smlDsServerSetConnectCallback) is
never triggered, which could be the reason why smlDsSessionDispatch
didn't occur as expected.

P.S. I am using libsyncml-0.4.6 in my tests.

Michael Bell wrote:

> Hi Josh,
>
> I'm back with a working notebook and I remember that you have at minimum
> one open issue issue with moreData (actually I search for the according
> mails).
>
> Josh Guilfoyle wrote:
>> Is libsyncml (either in SVN or 0.4.6) expected to function correctly
>> as an HTTP client to a SyncML server, also implemented using
>> libsyncml? I can't seem to make this work based on the
>> syncml-http-client example, and furthermore, I can't see that the API
>> even exposes enough to make this work. For example, how would you
>> specify the sync anchors? And why does it always seem to hardcode the
>> session id of 1 instead of using some crytographically secure random
>> number?
>
> 1. syncml-http-client
>
> I don't know who coded the http client in libsyncml/tools but after I
> tried to fix it for a while I give it up. It is completely broken and I
> think it never worked.
>
> 2. Example code
>
> If you want to see an example use of the http client stuff then please
> look into the syncml plugin of OpenSync. This is tested with OCS. It is
> some kind of special because OpenSync acts as a master and so it needs
> two sessions.
>
> 3. According sync anchors:
>
> Please see function ds_client_init_sync_mode in file syncml_ds_client.c
> of OpenSync plugin syncml.
>
> Best regards
>
> Michael

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Bugs in libwbxml

2008-06-20T03:32:57Z

Hi,

just for your information ...

libsyncml uses libwbxml for the conversion of XML/WBXML. I discovered
some bugs in libwbxml which cause error messages like:

String Table generation disabled: can't encode Literal

You can visit the bug at Debian:

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=487217

The bug report includes a patch.

Best regards

Michael
--
_______________________________________________________________

Michael Bell Humboldt-Universitaet zu Berlin

Tel.: +49 (0)30-2093 2482 ZE Computer- und Medienservice
Fax: +49 (0)30-2093 2704 Unter den Linden 6
michael.bell@... D-10099 Berlin
_______________________________________________________________

X.509 CA Certificates / Wurzelzertifikate

https://pki.pca.dfn.de/hu-berlin-ca/cgi-bin/pub/pki?cmd=getStaticPage;name=index;id=2&RA_ID=0

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

smime.p7s (7K) Download Attachment

Re: test core dump on Solaris

2008-06-19T23:50:16Z

Christian,

> Dear all,
>
> on my Solaris build host I see a lot of core files from
> test processes.
>
The main reason for this core is that printf kind function with null
string will cause core dump on Solaris, But Linux works well for this
situation. you can define LD_PRELOAD as following before running this
test to avoid this core:

export LD_PRELOAD=/usr/lib/0\@0.so.1

Jim

> I attached the output from pstack of the 4 newest cores.
> Anything else needed?
> Should I open a ticket?
> Shpuld we just ignore these core dumps?
>
> Affected processes are:
> continousbuild/build/tests/engine
> continousbuild/build/tests/error
> continousbuild/build/tests/ipc
> continousbuild/build/tests/updater
>
> Kind regards
>
> Christian
> --
> Christian Hilgers |ConSol*
> Tel. +49.2102.994-423 |Consulting&Solutions Software GmbH
> Fax +49.2102.994-411 |Berliner Str. 101, 40880 Ratingen
> email: Christian.Hilgers@... |WWW: http://www.consol.de
>
>------------------------------------------------------------------------
>
>-------------------------------------------------------------------------
>Check out the new SourceForge.net Marketplace.
>It's the best place to buy or sell services for
>just about anything Open Source.
>http://sourceforge.net/services/buy/index.php
>
>
>------------------------------------------------------------------------
>
>_______________________________________________
>Opensync-devel mailing list
>Opensync-devel@...
>https://lists.sourceforge.net/lists/listinfo/opensync-devel
>
>

Re: osync_xmlformat_validate() in format plugins

2008-06-19T07:18:02Z

Hi Daniel,

I started coding to get rid of osync_xml_validate and to have only one
instance of a xml schema for each objtype. Therefore I tried to reuse
the oop singleton pattern in C via a static GList. Could you please
review my first intuition because I am more familiar with C++ than C.

best regards,
Bjoern

> Hi,
>
> i just found that osync_xmlformat_validate() is quite expensive, since it
> parses for each call the XSLT schemas from scratch (xmlSchemaParse()).
>
> Callgraph of sync with vformat plugin and the vcard_to_xmlformat converter:
> http://cryptomilch.de/~dgollub/OpenSync/opensync_profiling_xmlformat_validate.png
> (Look for the redline in on the left bottom of the callgraph - 991x stands for
> the 991 function calls.)
>
> Actually we could do this XSLT schema parsing within in the OpenSync framework
> once, and validate all xmlformats which got reproted with the same object
> from a single xmlSchemaParse() call. This would require some changes for
> osync_xmlformat_*() - i try to prepare a patch within the next days. This
> would save us 990 of 991 xmlSchemaParse() calls in mentioned example. One
> call for each used xmlformat. If someone would sync contacts and events this
> would end up in two xmlSchemParse() calls for xmlformat-event and
> xmlformat-contact....
>
> Anyway- for now i would recommend to remove osync_xmlformat_validate() calls
> from every format plugin converter function. It's O.K. for debugging tools
> like "vconvert", but the plugin format conversion routines should be fast ;)
>
> I already removed osync_xmlformat_validate() from the vformat plugin and from
> gnokii-format and palm-format plugins. AFAIK there is only opie-sync and
> sync4j-sync (is anyone still using this?) left which have
> osync_xmlformat_validate(), at least in the OpenSync SVN - didn't checked yet
> other format plugins.
>
> best regards,
> Daniel
>

--
/Bjoern Ricks

Index: opensync_xmlformat.c
===================================================================
--- opensync_xmlformat.c (Revision 3370)
+++ opensync_xmlformat.c (Arbeitskopie)
@@ -91,6 +91,80 @@
return p;
}

+OSyncXMLFormatSchema * osync_xmlformat_schema_get_instance(OSyncXMLFormat * xmlformat, OSyncError **error) {
+ static GList * schemas = NULL;
+ OSyncXMLFormatSchema * osyncschema = NULL;
+ GList * entry;
+ const char * objtype;
+
+ osync_trace(TRACE_ENTRY, "%s(%p, %p)", __func__, xmlformat, error);
+
+ osync_assert(xmlformat);
+
+ objtype = osync_xmlformat_get_objtype(xmlformat);
+ //TODO get mutex
+ // find schema for objtype
+ for ( entry = schemas; entry != NULL; entry=entry->next) { // should be fast enough for only a few objtypes
+ osyncschema = (OSyncXMLFormatSchema *) entry->data;
+ if (!strcmp(osyncschema->objtype, objtype) ) {
+ return osyncschema;
+ }
+ osyncschema = NULL;
+ }
+ if ( osyncschema == NULL ) {
+ osyncschema = osync_try_malloc0(sizeof(OSyncXMLFormatSchema), error);
+ if(!osyncschema) {
+ osync_trace(TRACE_EXIT_ERROR, "%s: %s" , __func__, osync_error_print(error));
+ //TODO release mutex
+ return NULL;
+ }
+ osyncschema->objtype = g_strdup(objtype);
+ char *schemafilepath = g_strdup_printf("%s%c%s%s%s",
+ OPENSYNC_SCHEMASDIR,
+ G_DIR_SEPARATOR,
+ "xmlformat-",
+ osyncschema->objtype,
+ ".xsd");
+
+ xmlSchemaParserCtxtPtr xmlSchemaParserCtxt;
+ xmlSchemaPtr xmlSchema;
+ xmlSchemaValidCtxtPtr xmlSchemaValidCtxt;
+
+ xmlSchemaParserCtxt = xmlSchemaNewParserCtxt(schemafilepath);
+ osyncschema->schema = xmlSchemaParse(xmlSchemaParserCtxt);
+ xmlSchemaFreeParserCtxt(xmlSchemaParserCtxt);
+
+ osyncschema->context = xmlSchemaNewValidCtxt(xmlSchema);
+ if (osyncschema->context == NULL) {
+ xmlSchemaFree(osyncschema->schema);
+ g_free(osyncschema->objtype);
+ g_free(osyncschema);
+ osyncschema = NULL;
+ //TODO set error
+ }
+ schemas = g_list_append(schemas, osyncschema);
+ }
+ //TODO release mutex
+ return osyncschema;
+
+}
+
+
+osync_bool osync_xmlformat_schema_validate(OSyncXMLFormatSchema * schema, OSyncXMLFormat * xmlformat)
+{
+ osync_assert(xmlformat);
+ osync_assert(schema);
+
+ int rc = 0;
+
+ /* Validate the document */
+ rc = xmlSchemaValidateDoc(schema->context, xmlformat->doc);
+
+ if(rc != 0)
+ return FALSE;
+ return TRUE;
+}
+
/**
* @brief Validate the xmlformat against its schema in inidivual path
* @param xmlformat The pointer to a xmlformat object
@@ -399,8 +473,9 @@
osync_bool osync_xmlformat_validate(OSyncXMLFormat *xmlformat)
{
osync_assert(xmlformat);
-
- return _osync_xmlformat_validate(xmlformat, NULL);
+
+ OSyncXMLFormatSchema * schema = osync_xmlformat_schema_get_instance(xmlformat, NULL);
+ return osync_xmlformat_schema_validate(schema, xmlformat);
}

/**
Index: opensync_xmlformat_internals.h
===================================================================
--- opensync_xmlformat_internals.h (Revision 3370)
+++ opensync_xmlformat_internals.h (Arbeitskopie)
@@ -45,7 +45,20 @@

};

+struct OSyncXMLFormatSchema {
+
+ xmlSchemaPtr schema;
+
+ xmlSchemaValidCtxtPtr context;
+
+ const char *objtype
+};
+
int _osync_xmlformat_get_points(OSyncXMLPoints points[], int* cur_pos, int basic_points, const char* fieldname);
osync_bool _osync_xmlformat_validate(OSyncXMLFormat *xmlformat, const char *path);

+OSyncXMLFormatSchema * osync_xmlformat_schema_get_instance(OSyncXMLFormat * xmlformat, OSyncError **error)
+void osync_xmlformat_schema_free_instance(OSyncXMLFormatSchema * schema, OSyncError **error);
+osync_bool osync_xmlformat_schema_validate(OSyncXMLFormatSchema * schema, OSyncXMLFormat * xmlformat);
+
#endif /*OPENSYNC_XMLFORMAT_INTERNAL_H_*/

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

Re: test core dump on Solaris

2008-06-19T05:19:46Z

Jim is looking at these issues. There are some things cause Solaris fail
but Linux okay.

- shell
Linux is bash by default, solaris is sh
- diff
- named pipe

Thanks,
Halton.

On Thu, 2008-06-19 at 13:53 +0200, Christian Hilgers wrote:

> Dear all,
>
> on my Solaris build host I see a lot of core files from
> test processes.
>
> I attached the output from pstack of the 4 newest cores.
> Anything else needed?
> Should I open a ticket?
> Shpuld we just ignore these core dumps?
>
> Affected processes are:
> continousbuild/build/tests/engine
> continousbuild/build/tests/error
> continousbuild/build/tests/ipc
> continousbuild/build/tests/updater
>
> Kind regards
>
> Christian
> --
> Christian Hilgers |ConSol*
> Tel. +49.2102.994-423 |Consulting&Solutions Software GmbH
> Fax +49.2102.994-411 |Berliner Str. 101, 40880 Ratingen
> email: Christian.Hilgers@... |WWW: http://www.consol.de
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> Opensync-devel mailing list
> Opensync-devel@...
> https://lists.sourceforge.net/lists/listinfo/opensync-devel

test core dump on Solaris

2008-06-19T04:53:05Z

Dear all,

on my Solaris build host I see a lot of core files from
test processes.

I attached the output from pstack of the 4 newest cores.
Anything else needed?
Should I open a ticket?
Shpuld we just ignore these core dumps?

Affected processes are:
continousbuild/build/tests/engine
continousbuild/build/tests/error
continousbuild/build/tests/ipc
continousbuild/build/tests/updater

Kind regards

Christian
--
Christian Hilgers |ConSol*
Tel. +49.2102.994-423 |Consulting&Solutions Software GmbH
Fax +49.2102.994-411 |Berliner Str. 101, 40880 Ratingen
email: Christian.Hilgers@... |WWW: http://www.consol.de

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel

pstack.out.gz (2K) Download Attachment

Re: Patch for unit test failure and error only on Solaris

2008-06-18T07:02:09Z

Daniel:

>> +/*! @brief Copy files
>> + *
>> + * @param source source filename
>> + * @param dest destination filename
>> + * @returns TRUE on success, FALSE otherwise
>> + *
>> + */
>> +osync_bool osync_testing_file_copy(const char *source, const char *dest)
>> +{
>> + gchar *cmd;
>> + int ret;
>> +
>> + cmd = g_strdup_printf ("cp %s %s", source, dest);
>> + ret = system (cmd);
>> + g_free (cmd);
>>
>
> (Is there a certain reason for putting space between the function name and the
> arguments list? Not quite sure if this declared in the CODING file either...)
>
>

Not really! I'll modify this before commitment!

Thanks

Jim

>> +
>> + return ret;
>>
>
>
>
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> Opensync-devel mailing list
> Opensync-devel@...
> https://lists.sourceforge.net/lists/listinfo/opensync-devel
>

Re: Patch for unit test failure and error only on Solaris

2008-06-18T03:23:25Z

On Wednesday 18 June 2008 11:46:48 Jim Li wrote:
> I rewrite the patch according to your great suggestion. please help me
> to review it again.
>
[...]

Looks, good to me. We may replace the osync_testing_file_copy() code with
something more portable (and secure?) later. Feel free to commit.

> +/*! @brief Copy files
> + *
> + * @param source source filename
> + * @param dest destination filename
> + * @returns TRUE on success, FALSE otherwise
> + *
> + */
> +osync_bool osync_testing_file_copy(const char *source, const char *dest)
> +{
> +       gchar *cmd;
> +       int ret;
> +
> +       cmd = g_strdup_printf ("cp %s %s", source, dest);
> +       ret = system (cmd);
> +       g_free (cmd);

(Is there a certain reason for putting space between the function name and the
arguments list? Not quite sure if this declared in the CODING file either...)

> +
> + return ret;

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Opensync-devel mailing list
Opensync-devel@...
https://lists.sourceforge.net/lists/listinfo/opensync-devel