On Sun, Jul 05, 2015 at 07:28:34PM +0100, Guillaume M-M wrote: > Le 05/07/2015 18:06, Enrico Forestieri a écrit : > > > >Thanks for the reproducers. They evidence that the problem is due to > >the fact that MathData::metrics is not called when instant preview is > >on. I had tried to solve this issue by trying to update the relevant > >pointers when needed, but it seems that this causes issues with respect > >to the cursor position in some circumstances. In the attached patch > >for stable no attempt is made to call Buffer::getMacro and instead > >the macro_ pointer is updated by letting it point to the last known > >good copy of the MacroData. In this way I cannot succeed in triggering > >the assertion anymore. Please, test it with your non-minimal example > >and report back whether it solves the issue. > > > > > Thanks, I can no longer trigger the segfault with your patch.
Ok, I committed the patch to master. Richard, I fear this is also needed for stable. > By the way, you probably noticed that the preview in > lyx-segfault-dataloss.lyx fails, because \mathscr is not defined. Yes, it seems that it is due to the fact that the math inset is inside a LyX note. It also occurs with 2.1.3, so this is not a regression. -- Enrico
