Peter Eisentraut <peter_e@gmx.net> writes:
> It looks like you are building in fortify mode? I tried that a while
> ago and got a few more warnings. Are we trying to be fortify clean, and
> if so, what is our approach?
We're definitely *not* fortify-clean, although maybe trying to become so
would be a good idea. From memory, what I have seen in past builds on
Red Hat systems is a lot of warnings about ignoring the return value
from fwrite() and related functions. Which might not be an unreasonable
thing to try to get rid of, though I think many of the cases are in
places where there's no useful recovery action to be taken anyway.
(Whatcha gonna do if you fail to write the postmaster log ... try to
log a complaint message?)
In any case I agree that anything we try to do about this should be a
system-wide effort not a one-line hack.
> Also, considering my recent complaint about various brittleness in the
> regression test driver, more well hidden ignorings of errors are not
> exactly my favorite solution.
+1
regards, tom lane