]> granicus.if.org Git - postgresql/commit
Does it help to wait before reattaching?
authorTom Lane <tgl@sss.pgh.pa.us>
Tue, 1 May 2018 00:09:31 +0000 (20:09 -0400)
committerTom Lane <tgl@sss.pgh.pa.us>
Tue, 1 May 2018 00:09:31 +0000 (20:09 -0400)
commit23078689a9921968ac0873b017be6e7f772f10bc
tree6fef1968de6ded86c972eb2e0a234ddd5bede4e1
parent73042b8d136fe985746dc762fcb2a4356460842d
Does it help to wait before reattaching?

Revert the map/unmap dance I tried in commit 73042b8d1; that helps
not at all.

Instead, speculate that the unwanted allocation is being done on
another thread, and thus timing variations explain the apparent
unpredictability.  Temporarily add a 1-second sleep before the
VirtualFree call, in hopes that any such other threads will
quiesce and not jog our elbow.

This is obviously not a desirable long-term fix, but as a means of
investigation it seems useful.

Discussion: https://postgr.es/m/25495.1524517820@sss.pgh.pa.us
src/backend/port/win32_shmem.c