Hello Brad,
On 29/12/2021 01:04, Brad Smith wrote:
Hi Alessandro,
It looks like the issue mentioned in the Debian
bug report was fixed a bit under 2 years ago.
The issue with this port was very similar to that bug report, but the
root cause identified when we initially imported cad/magic was a
different Clang's pre-processor behavior w.r.t. GCC's one in line
continuation symbol treatment.
I see that you added lang/python as dependency (very good catch!); my
understanding is that when python is found during the config step, a
pre-processing script is used instead of CPP (I didn't notice that in
the past...), so this should solve the issue.
I would add a note in the commit log, in order to remember this point!
Regarding your diff, please see my comments in-line.
Can you please test this to make sure it works
as expected?
Index: Makefile
===
RCS file: /home/cvs/ports/cad/magic/Makefile,v
retrieving revision 1.6
diff -u -p -u -p -r1.6 Makefile
--- Makefile6 Mar 2021 11:59:59 - 1.6
+++ Makefile28 Dec 2021 22:53:21 -
@@ -2,6 +2,7 @@
COMMENT = interactive system for VLSI circuit layouts
DISTNAME =magic-8.3.137
+REVISION = 0
EXTRACT_SUFX =.tgz
CATEGORIES = cad
@@ -11,16 +12,17 @@ MAINTAINER =Alessandro De Laurenzis
-WANTLIB = ${MODTK_WANTLIB} c cairo fontconfig freetype m GL GLU X11
+WANTLIB = ${MODTK_WANTLIB} GL GLU X11 Xext Xi Xmu c cairo fontconfig \
+ freetype m
MASTER_SITES = http://opencircuitdesign.com/magic/archive/
-# The cpp preprocessor is used to generate .tech files; clang's breaks
-# newlines: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=698760
-COMPILER = ports-gcc
+COMPILER = base-clang ports-gcc
COMPILER_LANGS = c
-MODULES = x11/tk
+MODULES = lang/python \
+ x11/tk
+MODPY_RUNDEP = No
LIB_DEPENDS = ${MODTK_LIB_DEPENDS} \
graphics/cairo
@@ -31,16 +33,16 @@ WRKCONF = ${WRKSRC}/scripts
CONFIGURE_STYLE = gnu
CONFIGURE_ARGS += --with-distdir=${PREFIX} \
--with-tcl=${MODTCL_LIBDIR} \
- --with-tk=${MODTK_LIBDIR} \
- --with-cairo=${LOCALBASE}/include \
Removing --with-cairo=${LOCALBASE}/include I see the following warnings
during configure step:
checking cairo/cairo.h presence... no
configure: WARNING: cairo/cairo.h: accepted by the compiler, rejected by the
preprocessor!
configure: WARNING: cairo/cairo.h: proceeding with the compiler's result
so I would like to preserve that line.
- --enable-cairo-offscreen
This breaks OpenGL support, we definitely need that option.
+ --with-tk=${MODTK_LIBDIR}
+CONFIGURE_ENV += LDFLAGS="-L${LOCALBASE}/lib -L${X11BASE}/lib"
+
+CFLAGS += -I${LOCALBASE}/include -I${X11BASE}/include
The Clang compile log is much more noisy, so - considering that
cad/magic codebase is very old - I would add also the following:
# Really really not C99
CFLAGS += -Wno-return-type -Wno-parentheses
USE_GMAKE = Yes
NO_TEST = Yes
post-patch:
- @${SUBST_CMD} ${WRKSRC}/scripts/configure
@${SUBST_CMD} ${WRKSRC}/utils/paths.h
post-install:
Index: patches/patch-scripts_configure
===
RCS file: /home/cvs/ports/cad/magic/patches/patch-scripts_configure,v
retrieving revision 1.3
diff -u -p -u -p -r1.3 patch-scripts_configure
--- patches/patch-scripts_configure 6 Mar 2021 11:59:59 - 1.3
+++ patches/patch-scripts_configure 28 Dec 2021 21:50:55 -
@@ -21,20 +21,19 @@ Index: scripts/configure
# Not available on all versions: check for include file.
ac_fn_c_check_header_mongrel "$LINENO" "dlfcn.h" "ac_cv_header_dlfcn_h"
"$ac_includes_default"
if test "x$ac_cv_header_dlfcn_h" = xyes; then :
-@@ -8050,6 +8050,12 @@ fi
+@@ -8050,6 +8050,11 @@ fi
fi
;;
+*-openbsd*)
+ SHLIB_CFLAGS="-fPIC"
+ LDDL_FLAGS="-shared ${LIB_SPEC}"
-+ CFLAGS="${CFLAGS} -I${X11BASE}/include"
+ ;;
+
*-nextstep*)
LDDL_FLAGS="-nostdlib -r"
;;
-@@ -9485,16 +9491,5 @@ else
+@@ -9485,16 +9490,5 @@ else
echo
fi
fi
I see that the line "BUILD_DEPENDS = devel/m4" got lost in one of the
recent updates (but it is still needed AFAICS), so I would re-add it.
With the above modifications (except the additional CFLAGS line, which I
consider optional), this is OK maintainer.
Please find attached a modified patch for your convenience. Adding
ports@ in case other developers wants to chime in.
Also, I'm currently testing a newer upstream version, but that would
require a bit more time.
All the best
--
Alessandro De Laurenzis
[mailto:jus...@atlantide.mooo.com]
Web: http://www.atlantide.mooo.com
LinkedIn: http:/
