Rotate, move, break link messes up image library image shape

[GoogleCodeExporter]

What steps will reproduce the problem?
1. IAN image library > Science/Research > Measuring secchi depth
2. Rotate significantly
3. Move significantly
4. Break link to reference element

What is the expected output? What do you see instead?
Should look the same, but instead it jumps position and paths in the image get 
deformed. 

Please use labels and text to provide additional information.
The amount of deformation seems to relate to the amount rotated/moved. It seems 
to happen with most image library images, but found this to be a good example. 
I am pretty sure this is new since the new <symbol> approach to imports from 
the image library.

Original issue reported on code.google.com by adrianbj...@gmail.com on 11 Oct 2010 at 8:52

[GoogleCodeExporter]

I believe this has been fixed in r1811, though note that currently the 
resulting element has its rotation reset (stays rotated, but the new angle is 
now considered 0). Would you consider that a problem? 

Original comment by adeve...@gmail.com on 20 Oct 2010 at 2:04

• Changed state: Started

[GoogleCodeExporter]

Sorry for the really long silence on this - on vacation.

I think the position jumping problem has been solved, but the path deformation 
still exists with the latest revision - see the screenshot attached.

As for the rotation angle being set to zero - I wonder if it might be an issue 
with re-using of <symbol> paths - if the initial instance gets rotated and then 
set to zero, how would additional instances (<use>) of a symbol work? Note that 
at the moment extra instances redefine the <symbol> - see Issue 738

Original comment by adrianbj...@gmail.com on 18 Nov 2010 at 4:13

Attachments:

• deformed_path.png

[GoogleCodeExporter]

Actually I don't think the position jumping has been solved either. Try 
Flora>Mangroves>Ceriops australis leaf then move, rotate and break link - it 
jumps position considerably still.

Original comment by adrianbj...@gmail.com on 19 Nov 2010 at 1:04

[GoogleCodeExporter]

Hmm, bah. Alright, back to the drawing board I guess. :) And welcome back!

Original comment by adeve...@gmail.com on 19 Nov 2010 at 1:16

[GoogleCodeExporter]

I did what you said in comment #3 and did not notice any position 
jumping...tried in FF3.6 and Chrome 8.

I did however notice a new bug I must have introduced by fixing the dupe symbol 
issue, as an entirely different symbol was imported when I went back to the lib 
to add another one...whoops!

Original comment by adeve...@gmail.com on 17 Dec 2010 at 9:31

[GoogleCodeExporter]

It looks like r1811 did fix the jumping after all - sorry for the confusion on 
that - not sure what I did :) 

The path deformation with the "measuring secchi" example still exists though. 

I like the new "different symbol" bug - that is a good one - maybe it is a 
feature and not a bug - the new "random" symbol" feature :)

Original comment by adrianbj...@gmail.com on 20 Dec 2010 at 4:26

[GoogleCodeExporter]

Okay, should work much better as of r1889. There's still a problem where the 
gradients aren't converted right and indeed the original symbol is affected 
when using multiple instances.

Original comment by adeve...@gmail.com on 20 Dec 2010 at 6:42

[GoogleCodeExporter]

Definitely looking much better - thanks :)

Original comment by adrianbj...@gmail.com on 20 Dec 2010 at 7:29

[GoogleCodeExporter]

As of r1890 the original symbol will no longer be affected by a copy's 
rotate/break link effect. This by the way only occurred in Firefox because of 
its gradient elements existing outside the <symbol>.

Another note on your concern in comment #2: The rotation is only reset on the 
now unique element, so the original symbol is left untouched.

Now, the only remaining problem is the gradients changing on break-link...which 
is the same bug that occurs on path reorient and convert-to-path. 

Original comment by adeve...@gmail.com on 20 Dec 2010 at 8:08

[GoogleCodeExporter]

Sadly the solution to previously mentioned bug is not the same...but at least 
now there's a workaround: If you unlink the element while NOT rotated, and THEN 
rotate the resulting group, the gradients will appear as expected. 

Must not be too impossible to fix this, but having trouble right now.

Original comment by adeve...@gmail.com on 21 Dec 2010 at 7:39