hi tom,
% ls -ald /Volumes/data drwxr-x--x 12 openmac wheel 408 Nov 27 15:25 /Volumes/data/
Ah-hah, yes that's undoubtedly it.
and, since i've cleared that up pgsql is, again, behaving. yay.
for posterity's -- and other OSX users' -- sake, that's:
for the relevant drives/mounts in /Volumes, you need:
perms >= 555 (i use)
group should include your pgsql unprivileged user
'default' ownership, i believe, is root:admin for mounts, hence your pgsql user
would need to be added as a user to grup:admin
if this *does* turn out to be the case, does it make sense to have the script check perms up through the physical/path of the cwd and report with a more specific/desciptive error if they're not at least 555 on the mount point?
I'm not in the business of reimplementing getcwd(); especially not to deal with a problem that will break many other things besides Postgres.
whoa! i wasn't suggesting that at all ... rather, perhaps, simply a perms check from within the initdb script and an 'informative' error to the end-user. if i were the 1st/only to see such an issue, i'd, of course, not even bother with the suggestion ... but, as we've seen, it's raised its ugly head elsewhere (google: 'getcwd initdb')
I'd suggest filing a bug against whatever tool you used to create /Volumes/data/, suggesting that they adopt a less brain-dead default permissions setting. Execute-only on a mount point is just silly.
i have a suspicion it has to do with my filesharing/automount system/setup/settings. i'd had a bunch o' headaches a fairly long while back w/ getting Samba, AFS & NFS to play nice together. i'm fairly sure perms were changed, and the mount points may have been (inadvertently?) munged. why is 'suddenly' changed in the last days, i have no flippin' idea as yet!
postgresql may have been, simply, the 1st (me <-- surprised) app to 'hit' the getcwd path sensitivity.
hence, i think i have to file a bug-report with myself ... =)
two points o' feedback:
(a) without your help/suggestions, finiding this would've taken me a *lot* longer (thx!), and
(b) if this is in the pgsql docs (should it be?), i simply missed, misread or misunderstood it.
cheers,
richard
---------------------------(end of broadcast)--------------------------- TIP 4: Don't 'kill -9' the postmaster