Re: toolchain/53618: bogus warnings from printf %u and ntohs with -O1

[Hal Murray <murray+net%ip-64-139-1-69.sjc.megapath.net@localhost>]

The following reply was made to PR toolchain/53618; it has been noted by GNATS.

From: Hal Murray <murray+net%ip-64-139-1-69.sjc.megapath.net@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: toolchain-manager%netbsd.org@localhost, gnats-admin%netbsd.org@localhost,
    netbsd-bugs%netbsd.org@localhost, murray+net%ip-64-139-1-69.sjc.megapath.net@localhost
Subject: Re: toolchain/53618: bogus warnings from printf %u and ntohs with -O1
Date: Wed, 19 Sep 2018 01:28:10 -0700

 martin%duskware.de@localhost said:
 >  This is how it traditionally worked. It does not even depend on
 >  optimization level, on big endian CPUs all ntoh* are no-ops. 
 
 Thanks, but I don't understand what you are trying to tell me.
 
 I'm running on X86 so the no-op case shouldn't be happening.  Even if it did 
 happen, the argument to ntohs is an unsigned so it shouldn't complain.
 
 There are 4 printfs.  The first is there to make sure that I've got the 
 command line switches set to generate this type of warning.  The second is 
 there because I was curious.
 
 The last 2 are the interesting ones.  There are no complaints without -O1, but 
 warnings with -O1.
 
 If I actually run the code, both produce the same output and it is what I 
 expect.
 Yyy 13
 Y13 13
 Foo 3328
 Bar 3328
 
 3328 == 0xD00
 
 
 coypu%sdf.org@localhost said:
 >  this is because the definition of ntohs is
 >  #define ntohs(x) (x)
 >  in sys/endian.h. 
 
 That's the BYTE_ORDER == BIG_ENDIAN case
 I'm running on X86 which is LITTLE_ENDIAN
 
 #define ntohs(x)        bswap16(__CAST(uint16_t, (x)))
 
 
 -- 
 These are my opinions.  I hate spam.