Split window plugin shows circle markers for markers created by plugins in the main window

[techee]

For instance, in the screenshot below, the split view shows a circle where the git-changebar shows a diff and these markers are life-added/removed as the code is edited.

I had a look at the code but I don't understand how it is possible that the split window's Scintilla gets affected by the marker change in the main window. From what I understand, the split window plugin shares the underlying Scintilla buffer with the main window so the changes made in one editor appear in the other editor. But this IMO shouldn't affect things like markers because the Scintilla instances are separate.

I kind of understand why it's the white circle - it's the marker number 0 and it's not set to the right value for the split window but I don't understand why it appears at all. I also tried adding some extra code to git-changebar to make sure that e.g. some Geany callback doesn't return the split window's Scintilla but it doesn't seem to be the case.

(This is not some recent regression - I tried it with Geany 1.38 and it behaves the same way.)

[elextr]

Do the change markers have to be set up in each Scintilla individually? If so it probably isn't being done on the split because the plugins don't know about its existence (as its a plugin itself).

[techee]

Do the change markers have to be set up in each Scintilla individually? If so it probably isn't being done on the split because the plugins don't know about its existence (as its a plugin itself).

This is what I would expect - but then the markers shouldn't appear at all. I would expect no markers at all in the splitview plugin.

[elextr]

Scintilla docs, "By default, all 32 markers are set to SC_MARK_CIRCLE with a black foreground and a white background."

It would appear that the marker is added to the shared buffer, but the rendering is in each Scintilla.

[techee]

It would appear that the marker is added to the shared buffer

Oh, I misread the code - I thought it was using SCI_GETCHARACTERPOINTER and that would be sharing just gchar *. But it's SCI_GETDOCPOINTER so it's the Scintilla's "document" which is shared and it probably stores the information about markers too.

So yeah, it's less mysterious now and also probably not something we can fix easily.

[elextr]

Yeah, makes sense, the markers are per line, and lines exist in the document, not the view.

So per line data has to exist in the document, and per line data is where the marker number is stored, which shows as a white circle for any marker not properly set up in the view, for example in the split window view.

I don't think there is much we can do about any artefacts of interfering plugins, there have been cases of plugins using overlapping hard coded marker numbers etc.