Re: s_lock.h default definitions are rather confused

On 2015-01-10 23:03:36 +0100, Andres Freund wrote:
> On 2015-01-10 16:09:42 -0500, Tom Lane wrote:
> > I've not tried to build HEAD on my HPPA dinosaur for awhile, but I did
> > just now, and I am presented with boatloads of this:
> > 
> > ../../../src/include/storage/s_lock.h:759: warning: `S_UNLOCK' redefined
> > ../../../src/include/storage/s_lock.h:679: warning: this is the location of the previous definition
> > 
> > which is not too surprising because the "default" definition at line 679
> > precedes the HPPA-specific one at line 759.
> 
> That's 0709b7ee72e4bc71ad07b7120acd117265ab51d0.
> 
> Not too surprising that it broke and wasn't noticed without access to
> hppa - the hppa code uses gcc inline assembly outside of the big
> defined(__GNUC__) and inside the section headed "Platforms that use
> non-gcc inline assembly".
> 
> > I'm not particularly interested in untangling the effects of the recent
> > hackery in s_lock.h enough to figure out how the overall structure got
> > broken, but I trust one of you will clean up the mess.
> 
> I think it's easiest solved by moving the gcc inline assembly up to the
> rest of the gcc inline assembly. That'll require duplicating a couple
> lines, but seems easier to understand nonetheless. Not pretty.

Robert, do you have a better idea?

Alternatively we could just add a #undef in the hppa gcc section and
live with the uglyness.

Greetings,

Andres Freund

-- Andres Freund                       http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services