I tried to search whether this bug was reported in the past. Maybe it's the same as issue discussed here (plus I found a few more similar discussions), but because it's more than a year old discussion and the bug still exists, I wrote this post.
The problem was discovered by by customer, but original data cannot be shared, so I created very simple test case to illustrate the bug. Unfortunately, it's critical issue, blocking migration to CE (they use MicroStation and ORD). I informed my other customers working with similar data (literally, all civil customers) about this threat, because data are displayed differently.
There is 2D model, referencing another 2D DGN and DWG (3D, but with 2D elements at Z=0 in tested case). In the root model, there is a raster attached (tested with WMS, but raster file attached from disk has the same problem). Because the raster is at background plane, in V8i is correctly displayed behind all elements. But in CE U15, it hides elements from DWG:
I guess the problem itself is clear: Even when the raster has to be always at background, it's not. And because the results are different in CE U15, it's very dangerous (and migration blocking) bug, because some geometry can be hidden without any notification. And there is no chance, when typical civil design consists from tens of DGN and DWG files, to check every file whether everything is correct or not.
Quite accidentally I found a dependency: When references Update sequence dialog is used and root model is changed to be first, hidden elements are displayed. But it cannot be used as a workaround in many situations, because such change would break something else.
I will create also Service Ticket for this issue.
The test case is attached (used WMS is public). Can somebody from Bentley confirm the issue, whether it's defect #1064558 or a new one, and if its fix is planned for U16?
well, as I wrote, this one is crucial for me, because identified as one from migration blockers by my customer (and is for a majority of all my customer, because they compose DGN + DWG + rasters often). So it made a sense to invest my time.
Also, as developer (and certified SW tester ;-) I think I know how to report an issue (text + data set) in a way it can be repeated and analyzed quickly. When any issue cannot be reproduced easily (e.g. because of lack of test data or it's not repeatable), it probably will not be solved.
Plus, fortunately, I was able to create the simplest test case, based on original customer's data, that I am not allowed to share ;-)
Bentley Accredited Developer: iTwin Platform - AssociateLabyrinth Technology | dev.notes() | cad.point
Pellet Pusher said:The Update Sequence is incorrect if you want the references to display on top of the raster.
In my opinion you do not understand the reported problem: It's not about whether Update sequence is correct or not, which, btw, is), but that design files created in V8i are displayed (and printed) differently in CE, which can easily lead to serious troubles in engineering practice.
Pellet Pusher said:Think of it this way
Even when you are right in general, this rule is not valid for the discussed situation, because when a raster is placed at background plane, it was defined (at least in V8i) it will be always behind vector elements. It's the reason why the background plane (and foreground plane) exists.
Pellet Pusher said:With your drawings, this setting was available to me sometimes and not other times.
They are clear rules defined, when planes can be used (are supported), so it's not "sometimes". The planes were introduced together with element priority concept and they made Update sequence obsolete in 2D completely (but still useful in some 3D scenarios).
Pellet Pusher said:Since the Update Sequence was wrong in your drawings
Again, you missed the point: Something that works fine in V8i does not work in CE.
The second point you miss is that assignment to background planes have the highest priority (controls how raster are displayed) over all other settings.
Thanks for the investigation. I tested the DGN you sent, if we move the active model to the top in the Update Sequence list, all these 3 files display correctly,
I agree with Pellet, which should be WAD. The different display between v8i and CE , I think , was caused by the different settings of below 3 configuration variables:
Which control the default update sequence of references and master model in the design model.
The help document:
thanks for your answer.
chris.wu said:I agree with Pellet, which should be WAD.
With all respect, I tend to disagree with this conclusion. A part of the problem is that when background/foreground planes were introduced, it was never clearly defined what does it mean in context of older update sequence concept.
But the planes were always presented like "always at background" and "always at foreground", without any condition or dependency on the update sequence, attached structure of references etc. And it's how it works in V8i "out of the box" without any variables settings or update sequence changes.
chris.wu said:The help document:
Sorry, but as long as the documentation does not provide clear rules how the update sequence and background/foreground planes fit together, the help is not the justification of the problem.
chris.wu said:I think , was caused by the different settings of below 3 configuration variables:
I will check it (will try later today), but right now I am pretty sure in both configurations, no these variables are used, so we are back in the original issue "the same configuration for V8i and CE, but different result". But as I wrote, I will check and confirm the variables setting to be sure ;-)
a short update: I am able to confirm that used workspaces have the mentioned variables set in the same way as is standard in MicroStation (0 for DESIGN, 1 for DRAWING and SHEET). So everything should be right.
I will try to repeat the process how I created the test case, because I did not change the update sequence and it was wrong, so something is missed (maybe used seed file?). It will require some time, I will keep you informed.