Tom Lane wrote:
> Alvaro Herrera <alvherre@commandprompt.com> writes:
> > Excerpts from Bruce Momjian's message of jue sep 01 11:04:33 -0300 2011:
> >> I have re-run the script and applied the result, again tested on BSD and
> >> Linux. I will monitor the buildfarm for possible failures.
>
> > I think anything of this sort should be tested on Windows too.
>
> Well, if Windows is broken we'll find out soon enough from the
> buildfarm. My recollection from the last go-round is that the pain
> points were in non-default #define options that Bruce hadn't tested
> and that no buildfarm critter exercised either, such as LOCK_DEBUG.
Ah, but this time I only removed includes for files I could compile with
all #if markers removed. It only got 5.8k diff lines out a possible
6.8k lines, but this seems like an acceptable cost for greater
reliability.
-- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB
http://enterprisedb.com
+ It's impossible for everything to be true. +