No image in watch window

[zoneminder]

Hi Peter,

The caching headers I was referring to are the HTTP headers that are sent out prior to the image stream to tell the browser to not 'remember' any previous images but to always assume they are new and 'fresh'.

In .16 they are produced in the following lines

<i>
fprintf( fd, "Expires: Mon, 26 Jul 1997 05:00:00 GMT\r\n" );
fprintf( fd, "Last-Modified: %s\r\n", date_string );
fprintf( fd, "Cache-Control: no-store, no-cache, must-revalidate\r\n" );
fprintf( fd, "Cache-Control: post-check=0, pre-check=0\r\n" );
fprintf( fd, "Pragma: no-cache\r\n");
</i>

whereas in .15 they were just

<i>
fprintf( fd, "Pragma: no-cache\r\n" );
fprintf( fd, "Cache-Control: no-cache\r\n" );
fprintf( fd, "Expires: Thu, 01 Dec 1994 16:00:00 GMT\r\n" );
</i>

As an experiment maybe you could try editing zm_monitor.cpp and changing the lines back, I think they at about line 1019. If you do that and recompile and re-install and see if that makes any difference.

Phil,

[carrthomas]

Hi Phil,

Thanks.

Phil, just Before I change anything…… , for more information, here are the final entries of a measurement made using the ethereal network tool logging localhost traffic at the exact time my Mozilla browser froze tonight (approx 146 seconds from starting):-

3571 146,404 998 localhost.localdomain TCP 33792 > http [ACK] Seq=1788868653 Ack=17996443883 Win=49428 Len=0

3572 146,66697 localhost.localdomain HTTP Continuation

3573 146,666273 localhost.localdomain TCP 33792 > http [ACK] Seq=1788868653 Ack=1799647487 Win=49428 Len=0

3574 146,667271 localhost.localdomain TCP 33792 > http [FIN, ACK] Seq=1788868653 Ack=1799647487 Win=49428 Len=0

3575 146,67078 localhost.localdomain HTTP Continuation

3574 146,670809 localhost.localdomain TCP 33792 > http [RST] Seq=1788868654 Ack=0 Win=0 Len=0


Peter.

[zoneminder]

Hi Peter,

I'm sort of familiar with network logging in general though not really ethereal but I'm not sure what, if anything, you are trying to illustrate in your log above. Are you saying that's right, or that it is wrong, or neither?

Cheers,

Phil,

[maron]

I have the same problem, I've tried this with both mozilla on linux and windows, using cambozola and without using cambozola, I tried changing the cache in zm_monitor.cpp but with no luck.

Now as I am writing this I have the longest undisturbed session so far, that is after I set the jpeg compression parameter for streaming images from 10 to 70. So it looks like zm_monitor.cpp is having trouble streaming when compression rate is high. So it actually might have sometihing to do with this setting and then possibly hardware, (I'm using a 600Mhz machine with 392MB memory and a bt878 card).

[maron]

Just tried adding another screen and the problem starts again. Now it doesn't even help putting the image compression level to 100 (perfect quality). It has some syndroms though, in the montage view all 3 screens freeze at the exactly same time. My CPU is in about 40% use so it should have plenty left for the compression.

Makes one start wondering if it would be wise to use the loopback for v4l to pass the streaming on to a streaming protocol handler, and the montage would then use for example the vlc client and server for the video streaming, or would this be to CPU intensive?

[maron]

After fiddling with settings and testing a few times more it seems like this is related to writing an alarm to disk. As soon as an alarm is triggered the stream freezes. As soon as zm is finished writing the event to disk it resumes again. Does zm read the streams from disk? Would it then be better to let zm write streams to a small temporary disk. I'm evaluating zm for use in a productivitiy environment where the stream will be constantly monitored, but we will still want to record events. We've used 2 systems from kodicom up until now but the client side for those don't run on linux and the prices are way too high.

[zoneminder]

Hi Maron,

The monitor streaming has no disk element in it. The zms cgi program just reads the last image from shared memory, JPEGises it (with libjpeg) and writes it out to stdout. You can run zms from the command line by setting the QUERY_STRING environment variable to the parameters it is passed from the web and then just typing 'zms'. It will write out a load of binary though so redirect the output to a file. It should just run forever, if not then there is a problem with zms. I have the problem with streams just pegging out on Mozilla, if I run a montage they all tend to stop at different times though and I'm not sure why. It will run fine for hours in IE with Cambozola. Have you tried using Cambozola in Mozilla?

Phil,

[maron]

Thanks for the tip, I hadn't tried cambozola extensively in mozilla, Just tried it now for a few hours and decided to build a production system, since all works very well. This is excellent work you have done here. I found one little glitch though when using cambozola, that is probably a bug in the cambozola client. Sometimes when opening the montage view (with 3 cameras) in mozilla I get the same picture on 2 of them, refreshing a few times seems to fix the problem. I believe this is in the cambozola code since the html source of the montage view shows that cambozola get's the correct parameters.

[carrthomas]

Hi Phil,

The image freezes solid in the monitor watch window and then disappears leaving a kite face icon in the top corner of the image box.(I am afraid nothing has changed for me since last December when I started this mail)

If I do a refresh, and I get another instance of ZMS and all is OK for another short time before it freezes.

I am comforted to hear you also have the problem with streams just pegging out on Mozilla.

I know I could use cambozola but I feel that there is a fix to use mozilla somehow, but I can't find it. (other video code I use is OK with Mozilla)


Any Ideas I could try that you may have to get this working would be really welcome.

Thanks

Peter Thomas

[zoneminder]

Hi Peter,

It may not be much help now, but the next version of ZM will have MPEG based videostreaming which will hopefully address some of these problems. There are a few issues yet to be resolved and the first release may not work for everyone but hopefully it will provide a framework for progress.

Phil,