Recent changes to patches

#16 input_uvc.so refuses to handle symlinks

Alno — Wed, 09 Mar 2016 21:50:35 -0000

See https://sourceforge.net/p/mjpg-streamer/patches/17/ - that's a real reason behind symlink handling problems. I successfully use device symlinks after extending devicename buffer.

Support longer device names

Alno — Wed, 09 Mar 2016 21:48:49 -0000

Current buffer size and name length restriction for device name (12 is a snprintf argument, so real max len is 11) is too short - just enough for names like /dev/video3, but it failes to store longer device names such as /dev/video10 or /dev/video_c270 which may be present in systems with many video devices or custom udev rules.

Also, the same restriction is a real reason for device symlink problems (as described in https://sourceforge.net/p/mjpg-streamer/patches/16/) as symlinks usually have longer names.

This patch extends device name buffer size and snprintf call to allow device names up to 31 character, which should be enough for most cases.

#14 Support 3.16 and 3.17 kernels & then some

Dustin Spicuzza — Sun, 28 Feb 2016 21:08:12 -0000

It seems like the most popular github fork (300+ stars, 150+ forks) is https://github.com/jacksonliam/mjpg-streamer ... seems like it would be good to integrate with that instead. It is missing some of the mjpg-streamer-experimental changes (past October 2012), but it has a lot of good changes too.

input_uvc.so refuses to handle symlinks

Matthew Saunier — Fri, 08 May 2015 08:58:28 -0000

If a symlink to a device is passed to input_uvc.so, mjpg_streamer does not handle it correctly and refuses to start. This is not desirable behavior, as udev rules may be in place to give persistent device names (in the form of symlinks) to multiple attached cameras.

To fix this, a call to canonicalize_file_name() is needed to resolve the link. I've attached a very minimal patch to do just that, and I am successfully able to compile and run with no warnings.

Failure cases are handled as well, as init_VideoIn dies gracefully if passed a null from canonicalize_file_name(), which can happen if the user specifies a nonexistent device. Since this would inevitably result in total failure anyway, I don't consider this to be undesirable behavior.

Patch for ev3dev linux

Andres Kaasik — Sun, 21 Dec 2014 12:33:25 -0000

This patch fixes two bugs:

Using NB_BUFFER instead of real number of buffers granted (vd->rb.count)
More info http://linuxtv.org/downloads/v4l-dvb-apis/vidioc-reqbufs.html
"When the ioctl is called with a pointer to this structure the driver will attempt to allocate the requested number of buffers and it stores the actual number allocated in the count field."
Using pcontext->videoIn->buf.bytesused after buffer has been requeued. It appeares that pcontext->videoIn->buf.bytesused value will be 0 after the buffer is requeued. One needs to copy both the buffer content and save buffer size.

Support 3.16 and 3.17 kernels & then some

Oliver Schinagl — Thu, 30 Oct 2014 07:23:51 -0000

I've imported your SVN tree over on github, maintaining authors as best as I could etc. and added some patches, most recently supporting 3.16 and 3.17 kernels.

https://github.com/oliv3r/mjpg-streamer/

Pulling this git tree and pushing it to a sourceforge git tree should be without problem ;)

I'll be adding more patches to clean up the code base and fix issues/improve memory usage.

Olliver

#9 timestamp patch - httpd.c.diff

johnsmith — Mon, 23 Jun 2014 09:10:20 -0000

Yes, exactly the same here.

#13 output_cgi plugin: output_http with cgi support

Márton Miklós — Sun, 27 Oct 2013 13:18:27 -0000

Hello,

Thank you very much for your contribution!
A basic CGI functionality had been already added in the experimental branch.
I have used a different approach, but I like your unbuffered stduot redirection.

I will do a code cleanup next weekend, include merging the experimental mature features back to the stable branch.

#13 output_cgi plugin: output_http with cgi support

Austin Haigh — Sun, 28 Jul 2013 22:29:01 -0000

I forgot to mention (is there any way to edit tickets?) that it executes any script in the folder (or a subdirectory of the folder) passed as the '-w' parameter, but it could be easily restricted to a subdirectory (like 'cgi-bin'). I'll probably add a '--cgi' parameter that defaults to '/cgi-bin' eventually.

output_cgi plugin: output_http with cgi support

Austin Haigh — Sun, 28 Jul 2013 22:24:38 -0000

I'm adding cgi to the output_http plugin so it can execute server-side scripts. Right now it's very basic: no extra environment variables are passed to the script (like GET and POST variables), only 1 server can run at a time (but it can run with multiple output_http servers), and there is minimal error checking. Also, if the script handles SIGALRM signals without exiting a script can hang forever (normally SIGALRM is used to kill scripts after a short timeout). That said, it's good enough for its intended purpose, which is to display sensor readings next to a webcam stream.

https://github.com/austinHaigh/mjpg-streamer-output_cgi
https://github.com/austinHaigh/mjpg-streamer-output_cgi/archive/master.zip