On Wed, 2010-02-24 at 11:58 -0800, Jeremy Huddleston wrote:

> On Feb 24, 2010, at 05:45, Gaetan Nadon wrote:
> 
> > On Wed, 2010-02-24 at 09:34 +0100, Michel Dänzer wrote:
> > 
> >> On Tue, 2010-02-23 at 11:11 -0800, Jeremy Huddleston wrote: 
> >>> On Feb 23, 2010, at 11:07, Keith Packard wrote:
> >>> 
> >>>> On Tue, 23 Feb 2010 13:37:18 -0500, Gaetan Nadon  
> >>>> <[email protected]> wrote:
> >>>> 
> >>>>> This patch will ensure the xserver continues to suppress the
> >>>>> optimization, based on strict aliasing rules, after the option
> >>>>> is removed from $CWARNFLAGS. There is no change in the object
> >>>>> code produced.
> >>>> 
> >>>> I don't think we need to allow any of the code in the X server to be
> >>>> 'optimized' in this fashion, so I don't see a need to allow for
> >>>> per-directory selection.
> >>> 
> >>> It would be helpful if I had the option to remove this flag from  
> >>> XQuartz.
> >> 
> >> Out of curiosity, what significant benefit have you measured from using
> >> -fstrict-aliasing?
> >> 
> >> 
> > 
> > Excellent question which I carefully want to avoid :-) 
> > 
> > The problem I want to solve is the following: someone added
> > -fnostrict-aliasing a long time ago and I don't know why. Then it got
> > copied to a number of libraries, then got included in a macro
> > (XORG_CWARNFLAGS), which then got included by over a hundred modules,
> > still not knowing why.
> > 
> > There are over 50 modules that are compiling with no warning flags at
> > all. I don't want to contribute to the spread of this option. The patch
> > is about *transferring* the option out of the macro back to the modules
> > (where the skills are) and let them decide if they want that option or
> > not.
> 
> Additionally, before being copied into XORG_CWARNFLAGS fairly recently, this 
> was only in a handful of modules:
> 
> libICE
> libSM
> libX11
> libXau
> libXfont
> libXft
> libXpm
> libXres
> xorg-server
> (I didn't check xf86-drivers)
> 
> So adding it across the board is rather far-reaching, IMO.  I believe it 
> should only be added back to the modules listed above.
> 

For the sole purpose of providing information:

Modules which had CWARNFLAGS added a year ago *and* did not have
-fno-strict-aliasing option prior to that:


        font=   "util"
        app=    "appres bdftopcf beforelight bitmap editres fonttosfnt 
fslsfonts fstobdf iceauth \
                ico listres luit mkcomposecache mkfontscale oclock rendercheck 
rgb rstart  sessreg setxkbmap \
                showfont smproxy viewres x11perf xauth xbacklight xbiff xcalc 
xclipboard xclock xcmsdb xconsole \
                xcursorgen xdbedizzy xdm xdpyinfo xdriinfo xedit xev xeyes 
xf86dga xfd xfontsel xfs xfsinfo \
                xgamma xgc xhost xinit xkbcomp xkbevd xkbprint xkbutils xkill 
xload xlogo xlsatoms xlsclients \
                xlsfonts xmag  xman xmessage xmh xmodmap xmore xprop xrandr 
xrdb xrefresh xscope xset xsetmode \
                xsetpointer xsetroot xsm xstdcmap xvidtune xvinfo xwd xwininfo 
xwud"
        lib="   libAppleWM      libdmx          libfontenc      libFS           
libpciaccess    libWindowsWM \
                libXaw          libXcomposite   libXcursor      libXdamage      
libXdmcp        libXext \
                libXfixes       libXi           libXinerama     libxkbfile      
libXmu          libXrandr \
                libXrender      libXScrnSaver   libXt           libXtst         
libXv           libXvMC \
                libXxf86dga"
        driver="xf86-input-acecad       xf86-input-aiptek       
xf86-input-evdev \
                xf86-input-joystick     xf86-input-keyboard     
xf86-input-mouse \
                xf86-input-wacom        xf86-video-cirrus       
xf86-video-mach64 \
                xf86-video-qxl          xf86-video-siliconmotion"



Modules which never had -fno-strict-aliasing nor any warning flags of
any kind and still don't today:


        app="mkfontdir scripts twm xditview xinput"
        lib="libXxf86vm"
        driver="xf86-input-synaptics    xf86-input-vmmouse \
                xf86-input-void         xf86-video-apm          xf86-video-ark \
                xf86-video-ast          xf86-video-ati          
xf86-video-chips \
                xf86-video-dummy        xf86-video-fbdev\
                xf86-video-glint        xf86-video-i128         xf86-video-i740 
\
                xf86-video-mga          xf86-video-neomagic     
xf86-video-newport \
                xf86-video-nv           xf86-video-r128 \
                xf86-video-rendition    xf86-video-s3           
xf86-video-s3virge \
                xf86-video-savage       xf86-video-sis          
xf86-video-sisusb \
                xf86-video-suncg14      xf86-video-suncg3       
xf86-video-suncg6 \
                xf86-video-sunffb       xf86-video-sunleo       
xf86-video-suntcx \
                xf86-video-tdfx         xf86-video-tga          
xf86-video-trident \
                xf86-video-tseng        xf86-video-v4l          xf86-video-vesa 
\
                xf86-video-vmware       xf86-video-voodoo       xf86-video-wsfb 
\
                xf86-video-xgi          xf86-video-xgixp "


Modules which had -fno-strict-aliasing for a long time (prior to
CWARNFLAGS era)


        lib="libICE libSM libX11 libXau libXfont libXft libXpm libXres"
        driver="xf86-video-intel xf86-video-geode xf86-video-radeonhd"
        xserver


My next task is to add warning flags to those modules who don't. If we
wish to preserve status quo (for aliasing), I can define a new variable,
leaving CWARNFLAGS intact so older versions of modules configured
against newer versions of macros aren't affected. 


Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
xorg-devel mailing list
[email protected]
http://lists.x.org/mailman/listinfo/xorg-devel

Reply via email to