With pmake (BSD make) on GNU/Linux, some failures
| FAIL: ansi6.test
| FAIL: ansi7.test
| FAIL: ansi9.test
| FAIL: ansi10.test
happen because the `./ansi2knr' program doesn't match the `ansi2knr'
rule, thus pmake chooses its internal rule to compile C sources:
| gcc -g -O2 -o ./ansi2knr ./ansi2knr.c
thus the $(DEFS) don't get picked up:
| ./ansi2knr.c:184: error: conflicting types for 'malloc'
| ./ansi2knr.c:184: error: conflicting types for 'malloc'
I think the patch below fixes that. The issue also happens on BSD
systems, except that there it just doesn't happen to turn into a test
failure due to less conflicting malloc declarations:
| ./ansi2knr.c:184: warning: conflicting types for built-in function 'malloc'
OK? Or do we give up on KnR anyway and remove ansi2knr and its support
code?
Cheers,
Ralf
* lib/am/ansi2knr.am (ansi2knr): Rename target as..
(./ansi2knr): ..this, for BSD make.
(%ANSI2KNR-DIR%/ansi2knr): Adjust.
Index: lib/am/ansi2knr.am
===================================================================
RCS file: /cvs/automake/automake/lib/am/ansi2knr.am,v
retrieving revision 1.6
diff -u -r1.6 ansi2knr.am
--- lib/am/ansi2knr.am 14 May 2005 20:28:51 -0000 1.6
+++ lib/am/ansi2knr.am 2 May 2006 18:10:28 -0000
@@ -29,14 +29,14 @@
ANSI2KNR = %ANSI2KNR-DIR%/ansi2knr
%ANSI2KNR-DIR%/ansi2knr:
- cd %ANSI2KNR-DIR% && $(MAKE) $(AM_MAKEFLAGS) ansi2knr
+ cd %ANSI2KNR-DIR% && $(MAKE) $(AM_MAKEFLAGS) ./ansi2knr
else !%?ANSI2KNR-DIR%
## Substitution from AM_C_PROTOTYPES. This makes it be built only when
## necessary.
ANSI2KNR = @ANSI2KNR@
-ansi2knr: ansi2knr.$(OBJEXT)
+./ansi2knr: ansi2knr.$(OBJEXT)
$(LINK) ansi2knr.$(OBJEXT) $(LIBS)
ansi2knr.$(OBJEXT): $(CONFIG_HEADER)
