build gdbserver with -Werror by default

build gdbserver with -Werror by default

[Pedro Alves]

I've now applied this to build gdbserver with -Werror by default (on gcc),
and the usual --disable-werror/--enable-werror commands.

Tested on by building on x86_86-unknown-linux-gnu with
and without --disable-werror, and observing that -Werror
is included passed to gcc appropriately.

This is mostly copied from gdb/configure.ac, and I brought
the warning flags GDB uses, except for the -Wno-FOO settings,
since gdbserver was already building with -Wall (hence, doesn't
need to disable specific warnings -Wall implies).

This may break the build on some targets, which just means
that the target specific code should be fixed to be warning
free.

-- 
Pedro Alves

2010-08-26  Pedro Alves  <pedro@codesourcery.com>

	* Makefile.in (WARN_CFLAGS): Get it from configure.
	(WERROR_CFLAGS): New.
	(INTERNAL_CFLAGS): Add WERROR_CFLAGS.
	* configure.ac: Introduce --enable-werror, which adds -Werror to
	the compiler command line.  Enabled by default.  Disable with
	--disable-werror.  Add -Wdeclaration-after-statement
	Wpointer-arith and -Wformat-nonliteral to warning flags.
	* configure: Regenerate.

---
 gdb/gdbserver/Makefile.in  |    5 +--
 gdb/gdbserver/configure    |   69 +++++++++++++++++++++++++++++++++++++++++++++
 gdb/gdbserver/configure.ac |   42 +++++++++++++++++++++++++++
 3 files changed, 114 insertions(+), 2 deletions(-)

Index: src/gdb/gdbserver/Makefile.in
===================================================================
--- src.orig/gdb/gdbserver/Makefile.in	2010-07-31 08:53:16.000000000 +0100
+++ src/gdb/gdbserver/Makefile.in	2010-08-26 18:08:58.000000000 +0100
@@ -86,14 +86,15 @@ INCLUDE_CFLAGS = -I. -I${srcdir} -I$(src
 GLOBAL_CFLAGS = ${MT_CFLAGS} ${MH_CFLAGS}
 #PROFILE_CFLAGS = -pg
 
-WARN_CFLAGS = -Wall
+WARN_CFLAGS = @WARN_CFLAGS@
+WERROR_CFLAGS = @WERROR_CFLAGS@
 
 # CFLAGS is specifically reserved for setting from the command line
 # when running make.  I.E.  "make CFLAGS=-Wmissing-prototypes".
 CFLAGS = @CFLAGS@
 
 # INTERNAL_CFLAGS is the aggregate of all other *CFLAGS macros.
-INTERNAL_CFLAGS =  $(WARN_CFLAGS) ${CFLAGS} ${GLOBAL_CFLAGS} \
+INTERNAL_CFLAGS =  $(WARN_CFLAGS) $(WERROR_CFLAGS) ${CFLAGS} ${GLOBAL_CFLAGS} \
 	${PROFILE_CFLAGS} ${INCLUDE_CFLAGS}
 
 # LDFLAGS is specifically reserved for setting from the command line
Index: src/gdb/gdbserver/configure.ac
===================================================================
--- src.orig/gdb/gdbserver/configure.ac	2010-07-19 15:47:22.000000000 +0100
+++ src/gdb/gdbserver/configure.ac	2010-08-26 18:55:45.000000000 +0100
@@ -93,6 +93,48 @@ fi
 AC_SUBST(ustlibs)
 AC_SUBST(ustinc)
 
+AC_ARG_ENABLE(werror,
+  AS_HELP_STRING([--enable-werror], [treat compile warnings as errors]),
+  [case "${enableval}" in
+     yes | y) ERROR_ON_WARNING="yes" ;;
+     no | n)  ERROR_ON_WARNING="no" ;;
+     *) AC_MSG_ERROR(bad value ${enableval} for --enable-werror) ;;
+   esac])
+
+# Enable -Werror by default when using gcc
+if test "${GCC}" = yes -a -z "${ERROR_ON_WARNING}" ; then
+    ERROR_ON_WARNING=yes
+fi
+
+WERROR_CFLAGS=""
+if test "${ERROR_ON_WARNING}" = yes ; then
+    WERROR_CFLAGS="-Werror"
+fi
+
+build_warnings="-Wall -Wdeclaration-after-statement -Wpointer-arith \
+-Wformat-nonliteral "
+
+WARN_CFLAGS=""
+if test "x$GCC" = xyes
+then
+    AC_MSG_CHECKING(compiler warning flags)
+    # Separate out the -Werror flag as some files just cannot be
+    # compiled with it enabled.
+    for w in ${build_warnings}; do
+	case $w in
+	-Werr*) WERROR_CFLAGS=-Werror ;;
+	*) # Check that GCC accepts it
+	    saved_CFLAGS="$CFLAGS"
+	    CFLAGS="$CFLAGS $w"
+	    AC_TRY_COMPILE([],[],WARN_CFLAGS="${WARN_CFLAGS} $w",)
+	    CFLAGS="$saved_CFLAGS"
+	esac
+    done
+    AC_MSG_RESULT(${WARN_CFLAGS} ${WERROR_CFLAGS})
+fi
+AC_SUBST(WARN_CFLAGS)
+AC_SUBST(WERROR_CFLAGS)
+
 dnl dladdr is glibc-specific.  It is used by thread-db.c but only for
 dnl debugging messages.  It lives in -ldl which is handled below so we don't
 dnl use AC_CHECK_LIB (or AC_SEARCH_LIBS) here.  Instead we just temporarily
Index: src/gdb/gdbserver/configure
===================================================================
--- src.orig/gdb/gdbserver/configure	2010-07-19 15:47:22.000000000 +0100
+++ src/gdb/gdbserver/configure	2010-08-26 18:55:49.000000000 +0100
@@ -601,6 +601,8 @@ RDYNAMIC
 REPORT_BUGS_TEXI
 REPORT_BUGS_TO
 PKGVERSION
+WERROR_CFLAGS
+WARN_CFLAGS
 ustinc
 ustlibs
 LIBOBJS
@@ -673,6 +675,7 @@ enable_option_checking
 with_ust
 with_ust_include
 with_ust_lib
+enable_werror
 with_pkgversion
 with_bugurl
 with_libthread_db
@@ -1300,6 +1303,12 @@ if test -n "$ac_init_help"; then
 
   cat <<\_ACEOF
 
+Optional Features:
+  --disable-option-checking  ignore unrecognized --enable/--with options
+  --disable-FEATURE       do not include FEATURE (same as --enable-FEATURE=no)
+  --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
+  --enable-werror         treat compile warnings as errors
+
 Optional Packages:
   --with-PACKAGE[=ARG]    use PACKAGE [ARG=yes]
   --without-PACKAGE       do not use PACKAGE (same as --with-PACKAGE=no)
@@ -3914,6 +3923,66 @@ fi
 
 
 
+# Check whether --enable-werror was given.
+if test "${enable_werror+set}" = set; then :
+  enableval=$enable_werror; case "${enableval}" in
+     yes | y) ERROR_ON_WARNING="yes" ;;
+     no | n)  ERROR_ON_WARNING="no" ;;
+     *) as_fn_error "bad value ${enableval} for --enable-werror" "$LINENO" 5 ;;
+   esac
+fi
+
+
+# Enable -Werror by default when using gcc
+if test "${GCC}" = yes -a -z "${ERROR_ON_WARNING}" ; then
+    ERROR_ON_WARNING=yes
+fi
+
+WERROR_CFLAGS=""
+if test "${ERROR_ON_WARNING}" = yes ; then
+    WERROR_CFLAGS="-Werror"
+fi
+
+build_warnings="-Wall -Wdeclaration-after-statement -Wpointer-arith \
+-Wformat-nonliteral "
+
+WARN_CFLAGS=""
+if test "x$GCC" = xyes
+then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: checking compiler warning flags" >&5
+$as_echo_n "checking compiler warning flags... " >&6; }
+    # Separate out the -Werror flag as some files just cannot be
+    # compiled with it enabled.
+    for w in ${build_warnings}; do
+	case $w in
+	-Werr*) WERROR_CFLAGS=-Werror ;;
+	*) # Check that GCC accepts it
+	    saved_CFLAGS="$CFLAGS"
+	    CFLAGS="$CFLAGS $w"
+	    cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+if ac_fn_c_try_compile "$LINENO"; then :
+  WARN_CFLAGS="${WARN_CFLAGS} $w"
+fi
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+	    CFLAGS="$saved_CFLAGS"
+	esac
+    done
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: ${WARN_CFLAGS} ${WERROR_CFLAGS}" >&5
+$as_echo "${WARN_CFLAGS} ${WERROR_CFLAGS}" >&6; }
+fi
+
+
+
 old_LIBS="$LIBS"
 LIBS="$LIBS -ldl"
 for ac_func in dladdr

RE: build gdbserver with -Werror by default

[Pierre Muller]

  Hi Pedro,

  Cygwin 1.7 fails on gdbserver:
gcc -c -Wall -Wdeclaration-after-statement -Wpointer-arith -Wformat-nonliteral -
Werror -gstabs+ -O0    -I. -I../../../src/gdb/gdbserver -I../../../src/gdb/gdbse
rver/../common -I../../../src/gdb/gdbserver/../regformats -I../../../src/gdb/gdb
server/../../include ../../../src/gdb/gdbserver/tracepoint.c
cc1: warnings being treated as errors
../../../src/gdb/gdbserver/tracepoint.c: In function `add_tracepoint_action':
../../../src/gdb/gdbserver/tracepoint.c:1870: error: array subscript has type `c
har'
make[3]: *** [tracepoint.o] Error 1

  isxdigit argument type is int, not char according to man page...
Adding an explicit typecast to int at that line
allows to go to compilation of win32-low.c, but there you
end up with warnings about deprecated functions...
  To fix that, you would need to use the same kind of macros
as in gdb/windows-nat.c

Pierre Muller
Pascal language support maintainer for GDB

Re: build gdbserver with -Werror by default

[Pedro Alves]

On Friday 27 August 2010 09:55:11, Pierre Muller wrote:
>   Hi Pedro,
> 
>   Cygwin 1.7 fails on gdbserver:
> gcc -c -Wall -Wdeclaration-after-statement -Wpointer-arith -Wformat-nonliteral -
> Werror -gstabs+ -O0    -I. -I../../../src/gdb/gdbserver -I../../../src/gdb/gdbse
> rver/../common -I../../../src/gdb/gdbserver/../regformats -I../../../src/gdb/gdb
> server/../../include ../../../src/gdb/gdbserver/tracepoint.c
> cc1: warnings being treated as errors
> ../../../src/gdb/gdbserver/tracepoint.c: In function `add_tracepoint_action':
> ../../../src/gdb/gdbserver/tracepoint.c:1870: error: array subscript has type `c
> har'
> make[3]: *** [tracepoint.o] Error 1
> 
>   isxdigit argument type is int, not char according to man page...
> Adding an explicit typecast to int at that line
> allows to go to compilation of win32-low.c, 

and it'd fail elsewhere if it kept going, most probably.  GDB has
a bunch of suspicious code like this as well (e.g., in remote.c), so it
ended up with -Wno-char-subscripts.  I didn't bring that warning over to
gdbserver, as I didn't see any warning when building for linux and
mingw32, so I thought it was unnecessary (it is probably gcc version
dependent whether that's part of -Wall).  Since this is certainly
not a new warning (it's the fact that it causes an error that it's new)
I've now added -Wno-char-subscripts to gdbserver as well...

> but there you
> end up with warnings about deprecated functions...
>   To fix that, you would need to use the same kind of macros
> as in gdb/windows-nat.c

Hmm, I have no clue about what deprecated warnings are these,
and what macros you're talking about.  Are these new warnings, or
it is just that they've been there before, but ignored?

-- 
Pedro Alves

2010-08-27  Pedro Alves  <pedro@codesourcery.com>

	* configure.ac (build_warnings): Add -Wno-char-subscripts.
	* configure: Regenerate.

---
 gdb/gdbserver/configure    |    2 +-
 gdb/gdbserver/configure.ac |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

Index: src/gdb/gdbserver/configure.ac
===================================================================
--- src.orig/gdb/gdbserver/configure.ac	2010-08-26 19:32:03.000000000 +0100
+++ src/gdb/gdbserver/configure.ac	2010-08-27 11:02:56.000000000 +0100
@@ -112,7 +112,7 @@ if test "${ERROR_ON_WARNING}" = yes ; th
 fi
 
 build_warnings="-Wall -Wdeclaration-after-statement -Wpointer-arith \
--Wformat-nonliteral "
+-Wformat-nonliteral -Wno-char-subscripts"
 
 WARN_CFLAGS=""
 if test "x$GCC" = xyes
Index: src/gdb/gdbserver/configure
===================================================================
--- src.orig/gdb/gdbserver/configure	2010-08-26 19:32:03.000000000 +0100
+++ src/gdb/gdbserver/configure	2010-08-27 11:03:45.000000000 +0100
@@ -3944,7 +3944,7 @@ if test "${ERROR_ON_WARNING}" = yes ; th
 fi
 
 build_warnings="-Wall -Wdeclaration-after-statement -Wpointer-arith \
--Wformat-nonliteral "
+-Wformat-nonliteral -Wno-char-subscripts"
 
 WARN_CFLAGS=""
 if test "x$GCC" = xyes

Re: build gdbserver with -Werror by default

[Daniel Jacobowitz]

On Fri, Aug 27, 2010 at 11:29:08AM +0100, Pedro Alves wrote:
> and it'd fail elsewhere if it kept going, most probably.  GDB has
> a bunch of suspicious code like this as well (e.g., in remote.c), so it
> ended up with -Wno-char-subscripts.  I didn't bring that warning over to
> gdbserver, as I didn't see any warning when building for linux and
> mingw32, so I thought it was unnecessary (it is probably gcc version
> dependent whether that's part of -Wall).  Since this is certainly
> not a new warning (it's the fact that it causes an error that it's new)
> I've now added -Wno-char-subscripts to gdbserver as well...

Having just run into this, not all implementations of the C library
arrange for isdigit et al to report a char-subscript error.  The GLIBC
version casts the argument to int; just recently, newlib started
deliberately issuing the error.

-- 
Daniel Jacobowitz
CodeSourcery

Re: build gdbserver with -Werror by default

[Pedro Alves]

On Friday 27 August 2010 15:50:05, Daniel Jacobowitz wrote:
> 
> On Fri, Aug 27, 2010 at 11:29:08AM +0100, Pedro Alves wrote:
> > and it'd fail elsewhere if it kept going, most probably.  GDB has
> > a bunch of suspicious code like this as well (e.g., in remote.c), so it
> > ended up with -Wno-char-subscripts.  I didn't bring that warning over to
> > gdbserver, as I didn't see any warning when building for linux and
> > mingw32, so I thought it was unnecessary (it is probably gcc version
> > dependent whether that's part of -Wall).  Since this is certainly
> > not a new warning (it's the fact that it causes an error that it's new)
> > I've now added -Wno-char-subscripts to gdbserver as well...
> 
> Having just run into this, not all implementations of the C library
> arrange for isdigit et al to report a char-subscript error.  The GLIBC
> version casts the argument to int; just recently, newlib started
> deliberately issuing the error.

Ah, right, I followed that newlib discussion, but didn't connect
the dots in cygwin-uses-newlib.

-- 
Pedro Alves