Stacking a datasource over another stacked data source corrupts rrd figure

[GoogleCodeExporter]

What steps will reproduce the problem?
1. create rrd figure as follows:
graphDef.area(....)
graphDef.stack(..)
graphDef.stack(...)

What is the expected output? What do you see instead?
datasources stacked on top of each other. I see first datasource as area 
wrapped around with legend colors of the stacked datasources.

What version of the product are you using? On what operating system?
rrd4j-2.2, Windows 7

Please provide any additional information below.
I reviewed issue-37, I don't have any NaN in my figure. However, I do have 
zeros.

Original issue reported on code.google.com by safwan.o...@gmail.com on 20 Jun 2013 at 4:06

Attachments:

• area_area_stack_area_stack_http_ONE_DAY.gif
• area_stack_stack_area_stack_http_ONE_DAY.gif

[GoogleCodeExporter]

I don't understand where is the problem.

Are you complaining that the area for 504 is not always visible ?

Without the data it's not possible to analyze the bug.

Original comment by fbacche...@gmail.com on 16 Aug 2013 at 7:27

[GoogleCodeExporter]

I have confirmed the same issue. 

I am attaching a self-contained demo of the bug, plus 
data dumps and plots.

The ghost-border occurs whether you store NaN values or not.
When the steps are narrow enough, the ghost line looks like
actual non-zero values for that data source have been recorded.

Original comment by cinodapi...@gmail.com on 8 Oct 2014 at 2:23

Attachments:

• Rrd4jBug.java
• test_false.png
• test_false.xml
• test_true.png
• test_true.xml

[GoogleCodeExporter]

BTW, my work info:

Al Rossi
Fermi National Accelerator Laboratory
Feynman Computing Center 229A
Mail Station:     369 (FCC 2W)
Email:               arossi@fnal.gov
Phone:             (630) 840-3023

Original comment by cinodapi...@gmail.com on 8 Oct 2014 at 2:29

[GoogleCodeExporter]

OS Scientific Linux Fermi release 6.3 (Ramsey)

Version 2.2 for rrd4j.

java version "1.7.0_51"
Java(TM) SE Runtime Environment (build 1.7.0_51-b13)
Java HotSpot(TM) 64-Bit Server VM (build 24.51-b03, mixed mode)

Original comment by cinodapi...@gmail.com on 8 Oct 2014 at 2:32

[GoogleCodeExporter]

Further examples, from real-time plotting on our system.

Note how expanding the plot by decreasing the time-span from 15 days
to 2 days and then to 1 day reveals that the green lines are really
phantom artifacts surrounding the actual color, but when squeezed
actually make it seem as though they were recording "active restore"
data.  The underlying XML dump shows no recorded non-zero values 
for "active restore" (second to the last data source in the list).

Original comment by cinodapi...@gmail.com on 14 Oct 2014 at 2:43

Attachments:

• w-stkendca14a-4.rrd.xml
• w-stkendca14a-4@15.png
• w-stkendca14a-4@2.png
• w-stkendca14a-4@1.png