Re: [HACKERS] Parallel bitmap heap scan

On Wed, Mar 8, 2017 at 12:53 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Thanks.  Sorry for the hassle; my compiler isn't as picky about this
>> as I would like, and apparently Dilip's isn't either.
>
> Might be interesting to see whether -O level affects it.  In principle,
> whether you get the warning should depend on how much the compiler has
> analyzed the logic flow ...

What I'm using is:

Configured with:
--prefix=/Applications/Xcode.app/Contents/Developer/usr
--with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin14.5.0
Thread model: posix

While I haven't experimented with this too extensively, my general
impression is that this thing is extremely tolerant of uninitialized
variables.  I just tried compiling nodeBitmapHeapscan.c with -Wall
-Werror and each of -O0, -O1, -O2, and -O3, and none of those produced
any warnings.  I've been reluctant to go to the hassle of installing a
different compiler...

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company