Am 23.04.2012 10:39, schrieb Joe Orton:
On Thu, Apr 19, 2012 at 11:54:00AM -0000, [email protected] wrote:
Author: fuankg
Date: Thu Apr 19 11:53:59 2012
New Revision: 1327907
URL: http://svn.apache.org/viewvc?rev=1327907&view=rev
...
--- httpd/httpd/trunk/server/Makefile.in (original)
+++ httpd/httpd/trunk/server/Makefile.in Thu Apr 19 11:53:59 2012
@@ -27,9 +27,14 @@ TARGETS = delete-exports $(LTLIBRARY_NAM
include $(top_builddir)/build/rules.mk
include $(top_srcdir)/build/library.mk
+ifdef CC_FOR_BUILD
+gen_test_char: gen_test_char.c
+ $(CC_FOR_BUILD) -DCROSS_COMPILE -o $@ $<
+else
I think this is GNU make specific, isn't it? Our makefiles are (were?)
supposed to work with non-GNU make and I presume people still rely on
that for *BSD (?).
I thought that I saw somewhere in configure that we check for gnu make,
but I may recall wrong.
Anyway, I think it would be the best way if we would compeltely
eleminate building gen_test_char with apr = always build the same way as
happens if -DCROSS_COMPILE is defined. I we would change this then we
could always build gen_test_char with CC_FOR_BUILD where CC_FOR_BUILD ==
CC if not cross-compiling. This would then remove the hack in
Makefile.in. I dont believe that we would face any different output from
gen_test_char on any of the supported platforms if we would go this way ...
Gün.
