Steps to reproduce.
1. In Calc, in B2, insert a shape
2. Set Anchor "To Cell, Resize with Cell"
3. right click -> Position and Size -> rotate by some angle
4. Close and re-open
1. Shape stays in same position
1. Shape moves on each save
Created attachment 141216 [details]
Square shape originally anchored to B2
This seems to happen when the shape is larger than the cell it's anchored to.
Also seen in 5.4.x, so not a regression from my changes.
No repo in LibreOffice 3.3.4. A file saved with 5.3+ is correctly centered. Subsequent save, close, re-open cycles do not result in the shape moving.
Created attachment 141413 [details]
tail of terminal output, daily Linux dbgutil bibisect
Working on debian-buster in the daily Linux dbgutil bibisect
repository, I see the bug came into LO somewhere in the 73 or so
commits to master:
commit date s-h
-------- ---------- --------
good 559d6c25 2018-04-06 6c737acc
bad bdd69368 2018-04-07 3ec490fc
For this, I deemed:
bad, upon opening file: rotated square is in C3, extending into C2
bad after save and reopen, and good upon opening file and after
save and reopen: rotated square is in B2, extending into C2.
Contrary to comment 3, I have not seen the bug in 126.96.36.199.
I am setting keyword bibisected.
Created attachment 141427 [details]
Comparison LibreOffice 6.0 and Master
At import time, the drawing is over B2 in 6.0.3 and over C3 in master
Regression introduced by:
author Samuel Mehrbrodt <Samuel.Mehrbrodt@cib.de> 2018-04-06 10:20:43 +0200
committer Samuel Mehrbrodt <Samuel.Mehrbrodt@cib.de> 2018-04-06 13:09:55 +0200
commit 89b671c4a4288f3058157da292b1275e5bfb8392 (patch)
parent 49a2ca259287173d86a303b5c75d25ce52a9e26c (diff)
tdf#116836 Don't move objects out of cell when shrinking cell
Bisected with: bibisect-linux64-6.1
Adding Cc: to Samuel Mehrbrodt
Created attachment 141497 [details]
screenshot of 188.8.131.52 vs 184.108.40.206
I agree with Samuel here. I can reproduce this in 5.4, 5.3, 4.3, but not in 3.3 or 3.6. So this regression occurred between 3.6 and 4.3
Use these steps to reproduce this bug:
1. Open attachment 141216 [details]
2. Compare location of rectangle with my screenshot. LO 3.6 is correct. 4.3 is wrong.
I think Xisco & Terrence bisected a different bug.
I'm going to put in a bibisect for the regression that occurred here:
Version 220.127.116.11 (Build ID: 527dba6f6e0cfbbc71bd6e7b88a52699bb48799)
Build ID: 1b3956717a60d6ac35b133d7b0a0f5eb55e9155
attachment 141497 [details] shows the issue that I was reporting.
The bug Xisco & Terrence identified is probably related to Bug 116931. Changing anchor type from "To Cell, Resize with Cell" to "To Cell" will workaround this bug.
It seems I hit the bullseye with 41max: https://cgit.freedesktop.org/libreoffice/core/commit/?id=2b1aa949539d2fcbb3d349be3c279996630d83fc
fdo#56276 - resize/reposition rotated shapes in a sensible way
Committer is retired.