[LON-CAPA-users] dynamic plots not generated

[Stuart Raeburn]

Hi Todd,

> This happens only on his personal laptop (it works fine for him on other
> machines), running Windows7.  It happens with firefox, chrome, and ie.
> It's clearly some kind of issue with the configuration on his machine.

/res/csm/csmphyslib/Mechanics/Kinematics/Graphical_Analysis/PosAndAccelFromVel1.problem

If you mouse over the location of one of the dynamic plots (which are  
failing to render) and right click, and use "Copy Image Location" you  
should have a URL of the form:

http://loncapa.mines.edu/adm/roles/cgi-bin/plot.png?file=username_domain_1359314655_14857241_plot.data

where username is the username and domain correspond to the identity  
of the user for whom the plot was generated, and the first digit is  
the unix time when the page was requested, and the second digit  
includes some additional randomization.

Internally within LON-CAPA, the contents of the file:
/home/httpd/perl/tmp/username_domain_1359314655_14857241_plot.data
are retrieved and then passed as an argument to a piped call to gnuplot.
A header is then printed to STDOUT

Content-type: image/png

and then the output from gnuplot is sent.

There could be a problem displaying png images if a third-party  
application had been installed, client-side, which had altered the  
file class and MIME settings for the PNG file type.

If there is an issue with display of .png files specifically, the user  
could try replacing .png with .gif in the URL and resubmit the  
request, e.g.,

http://loncapa.mines.edu/adm/roles/cgi-bin/plot.gif?file=username_domain_1359314655_14857241_plot.data

Since the images in question are generated based on the contents of  
the query string, this issue could also be related to removal of the  
query string from the URI.

A google search for issues displaying images with Windows7 found:

http://support.microsoft.com/kb/283807
http://support.microsoft.com/kb/942172

which do *not* seem to be directly applicable to this case.

Anyway, a couple of things to check server side would be:
(a) look in /etc/httpd/logs/access_log for a line containing:

"GET  
/cgi-bin/plot.png?file=username_domain_1359314655_14857241_plot.data  
HTTP/1.1"

(replacing username and domain) to confirm that the response code and  
file size (the two numbers at the end of the line) are as expected.

(b) look in /home/httpd/perl/tmp for
username_domain_1359314655_14857241_plot.data

(replacing username and domain) to see if the data file which is  
passed to gnuplot is as expected.


Something to check client side would be to:
-- paste the URL of the image into a separate window and try to  
display the URL.

If the browser is firefox, an add-on such as httpfox
https://addons.mozilla.org/en-US/firefox/addon/httpfox/
could be installed (browser restart required) to allow the response  
header to be viewed.

The response header should be something like:

Date:
Server: Apache/2.2.3 (CentOS)
Connection: close
Transfer-Encoding: chunked
Content-Type: image/png

200 OK

The httpfox add-on will also display the bytes transferred.

Lastly is it possible to save the contents of the image URL to the  
desktop and open with some other application?


Stuart Raeburn
LON-CAPA Academic Consortium


Quoting Todd Ruskell <todd.ruskell at gmail.com>:

> Hi all,
>
> Here's one I don't recall seeing before.  I have one student on whose
> computer *some* dynamically generated plots don't display.  What's more
> interesting is that on refreshing the page, a different set of dynamically
> generated plots are not displayed, but some of those that were originally
> not displayed, are.
>
> This happens only on his personal laptop (it works fine for him on other
> machines), running Windows7.  It happens with firefox, chrome, and ie.
> It's clearly some kind of issue with the configuration on his machine.  I
> know it's a shot in the dark, but if you all have any ideas of what we
> might try to look for, I'd appreciate it.
>
> The resource in question is
> /res/csmphyslib/Mechanics/Kinematics/Graphical_Analysis/PosAndAccelFromVel1.problem
>
> which we've used regularly without incident for the last three semesters.
>
> Thanks for any small leads you can provide.
>
> Todd