On Sun, 2022-01-30 at 12:39 +0000, crashtest wrote: > New crashtest update available There are a number of duplicate asserts that have lingered for a while, for example https://dev-builds.libreoffice.org/crashtest/59dc59bb6e50d6d42d9fd6b82a165069d6e4bdbf/backtraces/task841-core.backtrace.txt I was finally able to reproduce one by taking the document of https://bugs.documentfoundation.org/attachment.cgi?id=163550 interactively loading it, and saving as odt and then wait a few seconds and the assert fires. I think I may have identified where this began to happen, with commit 493a916a3113e877835c9bc7c93faef0d29f9a33 Date: Mon Feb 22 14:28:59 2021 +0100 tdf#140158 tdf#138598 tdf#140598 sw: fix sync of AS_CHAR textboxes the porfly.cxx change of that commit. I find that if I disable the line calling SetFormatAttr(aNewTxBxAnchor) then the assert doesn't fire. I presume something goes out of sync when setting an anchor to a different SwTextNode than it was originally and this leads to the later assert. What the right solution is though it totally opaque to me.