9:58:07jackdanielport-set-mirror-geometry had such invocation, that was probably my doing, but I did not bother checking git blame
9:58:48jackdanielthe results were usually correct, because these were "old" mirror dimensions used to decide, whether we should skip the roundtrip to the display server
9:59:30jackdanielso if we set the mirror, then it was set to the right values, it was just they were sometimes not updated (rarely as it seems, only for the horizontal scroll-bar during initial exposure - as far as I can tell)
10:02:31jackdanielthat said I don't know whether it affects your fork -- I think that your repositories are set as 'private' on github; I might have introduced this regression later on