Why is svn 14061 a good idea?

[Rob Landley]

On Saturday 09 September 2006 1:09 pm, Denis Vlasenko wrote:
> I am thoroughly confused now. What's the consensus about
> reverting/non-reverting?
> 
> [please send mails to me directly too folks, not thru list...]

I think we need a config option for it, although I'd rather avoid confronting 
the user with this in menuconfig.

It seems recently that we've started to need config options that are set by 
platform.h or the makefiles rather than by menuconfig.  (Ok, we've needed 
them for a while, but we might want to actually _do_ something about it.)

The traditional approach to this is "./configure; make; make install", and 
although I'm all for tradition I really can't stand autoconf.  I also think 
it's silly for the project to have both "./configure --use-blah" and 
menuconfig, so I'm leaning against a separate ./configure step.  (We 
configure via the kernel idiom, as does uClibc.  It's strange, but it's a 
known strange rather than a bespoke strage.) 

The other can of worms here is that we have way too many CHECK_CC and CHECK_LD 
invocations at the start of the make.  I have a 2ghz laptop and depending on 
how clean my cache is this sucker has taken upwards of 15 seconds to start 
doing anything after I type make.  That's just SAD, and it's even worse when 
it's "make menuconfig" or "make help" that I'm typing, which isn't even going 
to use the result.

Probably what we need is some kind of generated makefile that we include and 
rebuild if it isn't there (and which gets zapped by make distclean if you 
wanna rebuild it for some reason).  Cache all the check_cc stuff, and system 
tests for "have I got gnu extensions".

.platform.make perhaps?

Rob
-- 
Never bet against the cheap plastic solution.