[haiku-commits] Re: r40300 - haiku/trunk/data/develop
- From: Philippe Houdoin <philippe.houdoin@xxxxxxxxx>
- To: haiku-commits@xxxxxxxxxxxxx
- Date: Mon, 31 Jan 2011 11:36:05 +0100
>> # set the linker and linker flags
>> - LD := gcc
>> + LD ?= gcc
>
> Have you checked all cases and side effects of this modification before
> commiting? That breaks my builds with following message:
> ======
> ld -o "objects.x86-gcc4-debug/Tolmach" objects.x86-gcc4-debug/TolmachApp.o
> objects.x86-gcc4-debug/Preferences.o objects.x86-gcc4-debug/TolmachWin.o
> objects.x86-gcc4-debug/TolmachView.o objects.x86-gcc4-debug/PGBHandler.o
> objects.x86-gcc4-debug/PGBIndex.o -g -Xlinker -soname=_APP_ -L./ -lroot
> -lbe -ltextencoding -lstdc++ -llocale -llocalestub
> ld: cannot find -linker
> ======
Looks like only gcc frontend is currently supported as linker command.
To allow LD overload, the LDFLAGS must be accomodate to support it.
> By the way, the makefile-engine have at the top some samples of another way
> of redefining variables:
> ======
> 16 ifeq ($(NAME), )
> 17 NAME = NameThisApp
> 18 endif
> ======
>
> and, I found such mixture of styles not a good idea. :-(
Well, checking that a variable is already defined or that is value is
empty is not exactly the same.
The ?= form will not detect the case when NAME is already defined but
it's value is empty:
NAME='' make
The current code don't care about overloading value, it cares about...
about... non-emptiness. (is that even a word!?).
And the make syntax don't allow much way to check that.
Bye,
Philippe.
Other related posts:
