<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 TRANSITIONAL//EN"> <HTML> <HEAD> <META HTTP-EQUIV="Content-Type" CONTENT="text/html; CHARSET=UTF-8"> <META NAME="GENERATOR" CONTENT="GtkHTML/3.32.2"> </HEAD> <BODY> On Sat, 2011-12-10 at 11:50 -0800, Jeremy Huddleston wrote: <BLOCKQUOTE TYPE=CITE> <PRE> On Dec 10, 2011, at 11:21 AM, Alan Coopersmith wrote: > On 12/10/11 10:13, Jeremy Huddleston wrote: >> Re: -Winline specifically since you brought it up. If the intent is to make an inline function for performance reasons, and that function actually can't be inlined, I'd think we'd want to know. Specifically regarding xtrans, perhaps we should just make those vararg functions !inline. > > Except that it only tells you that it can't be inlined by that particular > compiler on that platform - not that a different gcc version couldn't do > it, or a different compiler (clang, Solaris Studio, Intel, etc.) couldn't. > > Also, some of the warnings it generates in the Xserver are simply "optimizer > decided it was unprofitable to inline at this particular call site", not > "you designed this function in a way the optimizer can never inline it > anywhere". > > As I said, if you're doing specific performance tuning, I can see turning it > on, but in general builds it seems more like noise drowning out more useful > warning messages. Ok, fair enough.  Let's turn it off then for the next release.  This is why I wanted it to sit for a while before releasing util-macros, but it's always the case that things come up right after the release ;) </PRE> </BLOCKQUOTE> Fine tuning of the content can be released as 1.16.1 easily. The public interface does not change so there is no risk of breaking anything. <BLOCKQUOTE TYPE=CITE> <PRE> Are there any others that you think are too noisy or not as helpful as I had hoped them to be? </PRE> </BLOCKQUOTE> <BR> </BODY> </HTML>