Re: [PATCH] Use append attribution to CFLAGS in wcsmbs

[Mike Frysinger <vapier at gentoo dot org>]

On 13 Mar 2017 16:03, Gabriel F. T. Gomes wrote:
> On Mon, 13 Mar 2017 16:37:43 +0000 Joseph Myers wrote:
> > On Sat, 11 Mar 2017, Mike Frysinger wrote:
> > > On 08 Mar 2017 13:47, Gabriel F. T. Gomes wrote:  
> > > > CFLAGS for several files in wcsmbs/Makefile are set without the append
> > > > operator ("+="), thus ignoring other attributions.  This patch changes
> > > > that so Makefiles in sysdeps can set extra compilation flags.  This is
> > > > being done in preparation for future float128 patches.  
> > > 
> > > i'm not against this, but it seems like we should be doing this
> > > everywhere ?  when i grep for CFLAGS-xxx.c, there's about 650
> > > using = and 250 using +=.  
> > 
> > See the discussion starting at 
> > <https://sourceware.org/ml/libc-alpha/2012-11/msg00798.html> (and 
> > <https://sourceware.org/ml/libc-alpha/2013-01/msg00247.html>).  Given that 
> > CFLAGS are the right way of handling a particular file (rather than 
> > diagnostic pragmas, #define, etc., directly in the source code), I think 
> > appending makes sense unless there is a specific reason for not doing so 
> > in a particular case (and working out if any existing cases are 
> > deliberately not appending may be tricky).
> > 
> 
> In the discussion, there seems to be some consensus as to using
> sysdep-CFLAGS-<filename> (and sysdep-CFLAGS += $(sysdep-CFLAGS-$(<F))),
> instead of CFLAGS (as mentioned in the third item of [1]).  As well
> as there seems to be consensus on using the append operator (as
> mentioned in [2])
> 
> (Please, let me know if my perception of consensus is wrong here)

rereading the threads again, it looks like we're all in agreement.
let's land a commit to change all of the "=" to "+=", and let's add
support for $(sysdep-CFLAGS-$(<F)).
-mike

Attachment: signature.asc
Description: Digital signature