Re: [PATCH] Suppress Clang 3.9 warnings

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Noah Misch <noah(at)leadboat(dot)com>
Cc: Aleksander Alekseev <a(dot)alekseev(at)postgrespro(dot)ru>, David Steele <david(at)pgmasters(dot)net>, PostgreSQL Developers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: [PATCH] Suppress Clang 3.9 warnings
Date: 2017-03-16 14:17:45
Message-ID: 8651.1489673865@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Noah Misch <noah(at)leadboat(dot)com> writes:
> On Wed, Mar 15, 2017 at 12:04:11PM +0300, Aleksander Alekseev wrote:
>> If it no too much trouble could you please explain what will happen
>> on such platforms?

> Both port.h and a system header will furnish a strlcpy() declaration. The #if
> you modified exists to avoid that, and your change would make it ineffective
> for Clang. This will have no symptoms, or it will elicit a warning.

The reason why this is bad is that port.h's declaration might be different
from the system headers'. That's not hypothetical; for example, on my
Mac laptop, strlcpy is declared

size_t strlcpy(char * restrict dst, const char * restrict src, size_t size);

whereas of course there's no "restrict" in port.h. To make matters worse,
it looks like strlcpy is actually a macro expanding to
'__builtin___strlcpy_chk'. And the compiler on this box *is* clang,
meaning the proposed patch would affect it. When I try it, I get
boatloads of errors (not warnings) like these:

In file included from ../../src/include/postgres_fe.h:25:
In file included from ../../src/include/c.h:1125:
../../src/include/port.h:403:15: error: expected parameter declarator
extern size_t strlcpy(char *dst, const char *src, size_t siz);
^
/usr/include/secure/_string.h:105:44: note: expanded from macro 'strlcpy'
__builtin___strlcpy_chk (dest, src, len, __darwin_obsz (dest))
^

../../src/include/port.h:403:15: error: conflicting types for '__builtin___strlcpy_chk'
/usr/include/secure/_string.h:105:3: note: expanded from macro 'strlcpy'

In short, if this were to get committed, it would get reverted within
minutes, because more than a few of us use Macs.

regards, tom lane

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Joe Conway 2017-03-16 14:27:11 Re: Changing references of password encryption to hashing
Previous Message Heikki Linnakangas 2017-03-16 13:52:45 Re: scram and \password