On Mon, May 6, 2019 at 5:15 PM Peter Geoghegan <p...@bowt.ie> wrote: > VACUUM asserts P_FIRSTDATAKEY(opaque) > PageGetMaxOffsetNumber(page) > within _bt_mark_page_halfdead(), but doesn't test that condition in > release builds. This means that the earliest modifications of the > right page, before the high key PageAddItem(), are enough to cause a > subsequent "failed to re-find parent key" failure in VACUUM. Merely > setting the sibling blocks in the right page special area is enough to > cause VACUUM to refuse to run.
To be clear, my point here was that this confirms what you said about PageGetTempPage() failing after _bt_getbuf() has initialized the buffer for the new right page -- that is not in itself a problem. However, practically any other change to the right page that might occur before an error is raised within _bt_split() is a problem -- not just adding a new item. (You were right about that, too.) -- Peter Geoghegan