Comment (by Roland Haas):

Thank you. I will give it a try and if this fixes the issue for me as well propose it for backporting to the release branch. Looking at the code in the branch the commit 412c8c06 "Carpet: Adapt math functions to recent flesh changes" of carpet probably will block a direct merge into master since those “recent flesh changes” never made it the flesh’s master since they break on QueenBee (see https://bitbucket.org/einsteintoolkit/tickets/issues/2407/cactus-math-work-arounds-break-carpetx#comment-57679212 in #2470),

This patch most likely makes Carpet fail (more) to compile when CARPET_DISABLE_BBOXSET2 is set. Given that BBOXSET1 is no longer used on any clusters by default and thus slowly goes out of sync with bboxset2 over time, do you want me to add code to bboxset1 to provide iterator() (should be trivial, namely returning *this ought to do) or plan to retire bboxset1?

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2426/g-10-possibly-miscompiles-parts-of-carpet