GTK errors?

BattelleV · Post by **BattelleV** » Fri Dec 07, 2007 12:36 pm

Hello,

I am trying to evaluate your product, we have VSS running on a Windows server and are using your Fedora client to try to acess the database.
When I try I always get the message "FILE TIMES DO NOT MATCH - INFORMATION WILL BE LOST!"
In addition, the application has had a segmentation soon after (within a minute).
Below is additional text from the last crash.

I have GTK 2.X installed as well as GLIB.

Any help would be appreciated

FILE TIMES DO NOT MATCH - INFORMATION WILL BE LOST!

(sos:6688): Gtk-CRITICAL **: gtk_tree_store_get_value: assertion `VALID_ITER (iter, tree_store)' failed

(sos:6688): GLib-GObject-WARNING **: gtype.c

type id `0' is invalid

(sos:6688): GLib-GObject-WARNING **: can't peek value table for type `<invalid>' which is not currently referenced
Segmentation fault

Post by **jclausius** » Fri Dec 07, 2007 3:39 pm

Let's try to see if this is a Client side vs. Server side issue.

Can you try to access sos.sourceoffsite.com on port 8888, using user donald password duck?

Post by **jclausius** » Fri Dec 07, 2007 3:39 pm

Also, what distro (version of Fedora) are you using w/ what version of SourceOffSite?

BattelleV · Post by **BattelleV** » Fri Dec 07, 2007 3:51 pm

I can access the sos.sourceoffsite.com on port 8888 using donald password duck.
A clarification, the software is working now but twice today it has mysteriously crashed with the GTK debug information. I am using Fedora release 7 (Moonshine) and Sourceoffsite version 4.2.

I have been getting a lot of segmentation faults; but the GTK debug information has only occurred twice.

The second time the GTK debug variety crash occurred was when I was trying to reconnect to the database.

[vimal@localhost bin]$ ./sos
FILE TIMES DO NOT MATCH - INFORMATION WILL BE LOST!
Segmentation fault
[vimal@localhost bin]$ ./sos
[vimal@localhost bin]$ ./xemacs
[vimal@localhost bin]$ ./emacs
[vimal@localhost bin]$ ./sos
FILE TIMES DO NOT MATCH - INFORMATION WILL BE LOST!

(sos:6688): Gtk-CRITICAL **: gtk_tree_store_get_value: assertion `VALID_ITER (iter, tree_store)' failed

(sos:6688): GLib-GObject-WARNING **: gtype.c

type id `0' is invalid

(sos:6688): GLib-GObject-WARNING **: can't peek value table for type `<invalid>' which is not currently referenced
Segmentation fault
[vimal@localhost bin]$
[vimal@localhost bin]$ ./sos
FILE TIMES DO NOT MATCH - INFORMATION WILL BE LOST!
[vimal@localhost bin]$ ./sos
[vimal@localhost bin]$ ./sos
[vimal@localhost bin]$ ./sos

(sos:7000): Gtk-CRITICAL **: gtk_tree_store_get_value: assertion `VALID_ITER (iter, tree_store)' failed

(sos:7000): GLib-GObject-WARNING **: gtype.c

type id `0' is invalid

(sos:7000): GLib-GObject-WARNING **: can't peek value table for type `<invalid>' which is not currently referenced
Segmentation fault
[vimal@localhost bin]$ ./sos

Post by **jclausius** » Fri Dec 07, 2007 3:56 pm

The crash you encountered was with a local demo, not the demo server, correct?

If you try some sample usage on the demo server, does it seg fault as well?

BattelleV · Post by **BattelleV** » Fri Dec 07, 2007 4:49 pm

both, local server and to the remote one.

Vimal

Post by **jclausius** » Sat Dec 08, 2007 9:28 am

I want to make sure of this...

The crash you are seeing what servers are you connecting to? The local SourceOffSite Server? The Demo SourceOffSite Server?

BattelleV · Post by **BattelleV** » Mon Dec 10, 2007 9:21 am

I had originally assumed that it was occurring in both cases. Further testing shows it occurs when connecting to the local except for one test case below.

I think the application had segmentation faults with the remote version in the following scenario. I had sucessfully connected to the remote server and then tried to disconnect or reconnect to our local server (which it couldn't connect to); in those instances the application would try to connect, fail then either exit or segmentation fault and exit.

BattelleV · Post by **BattelleV** » Mon Dec 10, 2007 9:31 am

I tried to a branch in a test project on our local server. Another GTK related failure occurred and the application exited. This was after I selected a folder, decided to share it, unchecked recursive, selected branch after share and gave it a new name.

[vimal@localhost bin]$ ./sos

(sos:21903): Gtk-CRITICAL **: gtk_widget_set_sensitive: assertion `GTK_IS_WIDGET (widget)' failed

(sos:21903): Gtk-CRITICAL **: gtk_widget_set_sensitive: assertion `GTK_IS_WIDGET (widget)' failed
sos: VermProtocolMessage.cpp:212: bool CVermProtocolMessage::WriteReceivedFile(int, long int): Assertion `len == m_bodySize' failed.
Aborted
[vimal@localhost bin]$

Post by **jclausius** » Mon Dec 10, 2007 10:01 am

So you were able to test it. And then you executed a share branch and it crashed? Is that correct?

Outside of the share branch, if you reconnect to your local server now are things working?

Post by **jclausius** » Mon Dec 10, 2007 10:37 am

I just tested this with Debian on the test server...

1) created SOS-TEST
2) Selected SOS-TEST and chose Source -> Share
3) Choose PT_test within the share dialog and clicked OK.

Can you try the same test on the demo server?

If this works, it could be related to something on your server. SOS uses a VSS Automation component to talk to the database. What version is installed in your test environment? See - http://support.sourcegear.com/viewtopic.php?t=1510

Also, there is a server side log in the SOS Server install directory. Do you see any errors in there.

BattelleV · Post by **BattelleV** » Mon Dec 10, 2007 12:35 pm

The version of ssapi.dll we are using is 8.0.5.0727.42 with VSS 2005.
The log.txt did have error messages; I have attached a copy; incidentally we have deleted all the messages in the log file and will attempt to recreate some of the errors to get a better log file.

Thanks for looking into this Jeff. I appreciate the support.

Post by **jclausius** » Mon Dec 10, 2007 3:05 pm

One thing I noticed is some of the temporary working files of SourceOffSite is being locked.

Is there an active anti-virus scanner working on the SOS Server? If so, can you remove any active scans of the SOS working directory? From the error log, something is causing SOS issues when interacting with the VSS database by using that "scratch pad" directory.

BattelleV · Post by **BattelleV** » Tue Dec 11, 2007 9:29 am

Jeff,

Excluding the directory from the virus scanner has definitely helped. Sourceoffsite seems faster and more stable (i.e no segmentation faults as of this morning).

Thank you for all your help, I will evaluate SourceOffSite for our needs. In the meantime can you answer a few questions for me or point me in the right direction to look?
What are the major differences (features/functionality) between SourceOffSite and VSS?
Do you have a current bug list for SourceOffSite that we can review?
We plan on doing most of our development in Linux, and maintaining our server on a Windows Box, I noticed in one of the other posts that labeling is not supported by SourceOffSite. How well do SourceOffSite and VSS function together? Is it possible to use VSS to open the database and do labelling etc? Is it advisable to do so?

Any insight into any of the above will be greatly appreciated.

Thanks

Vimal

lbauer · Post by **lbauer** » Tue Dec 11, 2007 11:05 am

We don't make our bug list for SOS public, sorry.

But some of the differences between SOS and VSS are that VSS does not have clients that work on Mac or Unix.

SOS shows the status of files in your working directory, so you can see if a file is modified, out of date, or not in the working folder.

VSS does not work as well as SOS over the Internet.

SOS *does* support labeling. It does not support label promotion, though. You would still need to do that with a VSS client. SOS also does not support file rollback or find in files. SOS does not have a merge branches feature.

Some of the differences in SOS and VSS are not bugs as much as limitation of the VSS automation component, which the SOS Server uses to communicate with the VSS database.