6:19 p.m.
PATCH 21.5
This patch captures the values of __GNUC__ and __GNUC_MINOR__ in
configure variables when gcc is used. It uses them to see if gcc 3.4 or
later is in use and, if so, adds -Wdeclaration-after-statement to the
warning flags when compiling with C.
I am interested in finding out whether this approach is acceptable to
the other developers. In particular, is the $(( ... )) syntax okay? It
is standard Bourne shell (see the Single Unix Specification on
www.unix.org), but it is not used anywhere in our present configure
script. Is anyone using a shell that does not support that syntax?
ChangeLog addition:
2006-06-20 Jerry James <james(a)xemacs.org>
* configure.ac: When gcc is used, capture the values of __GNUC__
and __GNUC_MINOR__. Use them to conditionally add warning flags
that are not supported by all versions of gcc.
xemacs-21.5 source patch:
Diff command: cvs -q diff -uN
Files affected: configure.ac
Index: configure.ac
===================================================================
RCS file: /pack/xemacscvs/XEmacs/xemacs/configure.ac,v
retrieving revision 1.43
diff -d -u -r1.43 configure.ac
--- configure.ac 2006/06/06 00:18:29 1.43
+++ configure.ac 2006/06/20 17:11:33
＠＠ -1883,15 +1883,13 ＠＠
dnl Determine GCC version.
if test "$GCC" = "yes"; then
AC_RUN_IFELSE([AC_LANG_SOURCE([int main () {
-#if __GNUC__ >= 3
-return 11;
-#else
-return 0;
-#endif
-}])], [:],
-[case "$?" in
- 11) echo "You appear to be using GCC version 3 or above."; __GCC3=yes ;;
-esac])
+return __GNUC__;
+}])], [:], [__GCC="$?"])
+AC_RUN_IFELSE([AC_LANG_SOURCE([int main () {
+return __GNUC_MINOR__;
+}])], [:], [__GCC_MINOR="$?"])
+__GCC_VERSION=$__GCC"."$__GCC_MINOR
+echo "You appear to be using GCC version" $__GCC_VERSION
fi
dnl Support for using a different compiler for xemacs itself.
＠＠ -2311,7 +2309,7 ＠＠
dnl Warnings about char subscripts are pretty pointless, though,
dnl and we use them in various places.
with_cflags_warning="$with_cflags_warning -Wsign-compare
-Wno-char-subscripts"
- test "$__GCC3" = "yes" &&
with_cflags_warning="$with_cflags_warning -Wpacked"
+ test $((__GCC)) -ge 3 && with_cflags_warning="$with_cflags_warning
-Wpacked"
dnl glibc is intentionally not `-Wpointer-arith'-clean.
dnl Ulrich Drepper has rejected patches to fix the glibc header files.
test "$have_glibc" != "yes" && \
＠＠ -2327,8 +2325,11 ＠＠
dnl some very ugly code changes.
if test "$with_xemacs_compiler" = "g++"; then
xe_cflags_warning="$with_cflags_warning -Weffc++"
- elif test "$__GCC3" = "yes"; then
+ elif test $((__GCC)) -ge 3; then
xe_cflags_warning="$with_cflags_warning -Wunused-parameter"
+ if test $((__GCC)) -gt 3 || test $((__GCC_MINOR)) -ge 4; then
+ with_cflags_warning="$with_cflags_warning
-Wdeclaration-after-statement"
+ fi
fi
with_cflags_warning="$with_cflags_warning -Wshadow -Wmissing-declarations"
with_cflags_warning="$with_cflags_warning -Wmissing-prototypes
-Wstrict-prototypes"
--
Jerry James, Assistant Professor james(a)xemacs.org
Computer Science Department http://www.cs.usu.edu/~jerry/
Utah State University
7:25 a.m.
>>>> "Jerry" == Jerry James
<james(a)xemacs.org> writes:
Jerry> I am interested in finding out whether this approach is
Jerry> acceptable to the other developers.
+1 on the approach.
Jerry> In particular, is the $(( ... )) syntax okay?
ash, which was the default /bin/sh on at least some Debian systems[1], is
buggy, at least in version 0.5.3-3.
if test $((__GCC)) -ge 3; then echo not smaller; else echo smaller; fi
gives 'ash: arith: syntax error: "__GCC"', while
if test $(($__GCC)) -ge 3; then echo not smaller; else echo smaller; fi
gives "not smaller". (Note the extra '$' in the second command.)
Blech.
I don't see a good reason to use the $((expr))syntax instead of
ordinary variable substitution. All of zsh 4.2.5, (Debian) ash
0.5.3-3, and bash 3.1.14(1)-release work with 'test "$__GCC" -ge 3'
for numerical values of __GCC. The conf programs presumably are
guaranteed to give numerical output for any usable version of gcc,
right?
Footnotes:
[1] At one point Debian was trying to flush out bash-isms. This may
no longer be relevant.
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
1:19 a.m.
New subject: [S 21.5] -Wdeclaration-after-statement support
SUPERCEDES 21.5
"Stephen J. Turnbull" <stephen(a)xemacs.org> wrote:
I don't see a good reason to use the $((expr))syntax instead of
ordinary variable substitution. All of zsh 4.2.5, (Debian) ash
0.5.3-3, and bash 3.1.14(1)-release work with 'test "$__GCC" -ge 3'
for numerical values of __GCC. The conf programs presumably are
guaranteed to give numerical output for any usable version of gcc,
right?
Sorry, I was just having a brain cramp in public. I tried
'test $__GCC -ge 3' (i.e., without the quotes around $__GCC) and when
that didn't work, I got the $(()) stuff out of the bash manual.
Oooooookay, time for me to go home and rest for awhile, isn't it?
I'm including a superceding patch below. I'm concerned though, because
a test run of configure produced a top-level Makefile that shows this:
CFLAGS=-Wall -Wno-switch -Wundef -Wsign-compare -Wno-char-subscripts -Wpacked
-Wdeclaration-after-statement -Wshadow -Wmissing-declarations -Wmissing-prototypes
-Wstrict-prototypes -g -O2
but the Installation file contains this line:
- Compiler command: gcc -Wall -Wno-switch -Wundef -Wsign-compare
-Wno-char-subscripts -Wpacked -Wunused-parameter -g -O2
which matches the CFLAGS setting in src/Makefile. Why is the top-level
Makefile different from src/Makefile in the setting of CFLAGS? Since
it's only mentioned in the top-level Makefile, am I really going to get
-Wdeclaration-after-statement? (Time passes while a quick test is
performed.)
No! I'm only getting -Wdeclaration-after-statement while compiling in
lib-src. The compilation in lwlib and src uses the CFLAGS reported in
Installation. What's going on here?
Index: configure.ac
===================================================================
RCS file: /pack/xemacscvs/XEmacs/xemacs/configure.ac,v
retrieving revision 1.43
diff -d -u -r1.43 configure.ac
--- configure.ac 2006/06/06 00:18:29 1.43
+++ configure.ac 2006/06/22 00:04:55
＠＠ -1883,15 +1883,13 ＠＠
dnl Determine GCC version.
if test "$GCC" = "yes"; then
AC_RUN_IFELSE([AC_LANG_SOURCE([int main () {
-#if __GNUC__ >= 3
-return 11;
-#else
-return 0;
-#endif
-}])], [:],
-[case "$?" in
- 11) echo "You appear to be using GCC version 3 or above."; __GCC3=yes ;;
-esac])
+return __GNUC__;
+}])], [:], [__GCC="$?"])
+AC_RUN_IFELSE([AC_LANG_SOURCE([int main () {
+return __GNUC_MINOR__;
+}])], [:], [__GCC_MINOR="$?"])
+__GCC_VERSION=$__GCC"."$__GCC_MINOR
+echo "You appear to be using GCC version" $__GCC_VERSION
fi
dnl Support for using a different compiler for xemacs itself.
＠＠ -2311,7 +2309,7 ＠＠
dnl Warnings about char subscripts are pretty pointless, though,
dnl and we use them in various places.
with_cflags_warning="$with_cflags_warning -Wsign-compare
-Wno-char-subscripts"
- test "$__GCC3" = "yes" &&
with_cflags_warning="$with_cflags_warning -Wpacked"
+ test "$__GCC" -ge 3 &&
with_cflags_warning="$with_cflags_warning -Wpacked"
dnl glibc is intentionally not `-Wpointer-arith'-clean.
dnl Ulrich Drepper has rejected patches to fix the glibc header files.
test "$have_glibc" != "yes" && \
＠＠ -2327,8 +2325,11 ＠＠
dnl some very ugly code changes.
if test "$with_xemacs_compiler" = "g++"; then
xe_cflags_warning="$with_cflags_warning -Weffc++"
- elif test "$__GCC3" = "yes"; then
+ elif test "$__GCC" -ge 3; then
xe_cflags_warning="$with_cflags_warning -Wunused-parameter"
+ if test "$__GCC" -gt 3 || test "$__GCC_MINOR" -ge 4; then
+ with_cflags_warning="$with_cflags_warning
-Wdeclaration-after-statement"
+ fi
fi
with_cflags_warning="$with_cflags_warning -Wshadow -Wmissing-declarations"
with_cflags_warning="$with_cflags_warning -Wmissing-prototypes
-Wstrict-prototypes"
--
Jerry James, Assistant Professor james(a)xemacs.org
Computer Science Department http://www.cs.usu.edu/~jerry/
Utah State University
11:15 a.m.
New subject: [S 21.5] -Wdeclaration-after-statement support
>>>> "Jerry" == Jerry James
<james(a)xemacs.org> writes:
Jerry> No! I'm only getting -Wdeclaration-after-statement while
Jerry> compiling in lib-src. The compilation in lwlib and src
Jerry> uses the CFLAGS reported in Installation. What's going on
Jerry> here?
--with-xemacs-compiler support. There's an xe_cflags_warning variable
that you need to propagate those options to. It needs to be in that.
Since g++ won't need it, you can just duplicate it in the place where
you're setting with_cflags_warning.
Something like this:
＠＠ -2327,8 +2325,12 ＠＠
dnl some very ugly code changes.
if test "$with_xemacs_compiler" = "g++"; then
xe_cflags_warning="$with_cflags_warning -Weffc++"
- elif test "$__GCC3" = "yes"; then
+ elif test "$__GCC" -ge 3; then
xe_cflags_warning="$with_cflags_warning -Wunused-parameter"
+ if test "$__GCC" -gt 3 || test "$__GCC_MINOR" -ge 4; then
+ with_cflags_warning="$with_cflags_warning
-Wdeclaration-after-statement"
+ xe_cflags_warning="$with_cflags_warning
-Wdeclaration-after-statement"
+ fi
fi
with_cflags_warning="$with_cflags_warning -Wshadow
-Wmissing-declarations"
with_cflags_warning="$with_cflags_warning -Wmissing-prototypes
-Wstrict-prototypes"
Urk. Note those last two lines are getting lost, too.
xe_cflags_warning versions probably should be added somewhere in the
above.
What do you think about redesigning the logic so that if with_compiler
!= with_xemacs_compiler two separate runs of the tests are done, with
their own values of __GCC and __GCC_MINOR?
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
6:05 p.m.
New subject: [S 21.5] -Wdeclaration-after-statement support
"Stephen J. Turnbull" <stephen(a)xemacs.org> wrote:
What do you think about redesigning the logic so that if
with_compiler
!= with_xemacs_compiler two separate runs of the tests are done, with
their own values of __GCC and __GCC_MINOR?
Yes, that sounds like a good idea. I'll try to get to it in the next
few days, unless somebody else wants to grab the glory.
--
Jerry James, Assistant Professor james(a)xemacs.org
Computer Science Department http://www.cs.usu.edu/~jerry/
Utah State University
2:33 p.m.
New subject: [S 21.5] -Wdeclaration-after-statement support
>>>> "Jerry" == Jerry James
<james(a)xemacs.org> writes:
Jerry> "Stephen J. Turnbull" <stephen(a)xemacs.org> wrote:
> What do you think about redesigning the logic so that if
> with_compiler != with_xemacs_compiler two separate runs of the
> tests are done, with their own values of __GCC and __GCC_MINOR?
Jerry> Yes, that sounds like a good idea. I'll try to get to it
Jerry> in the next few days, unless somebody else wants to grab
Jerry> the glory.
Please drop a line to xemacs-beta or xemacs-design if you start on
it. I want to review configure for Xft-related bugs, and I've worked
on the with_xemacs_compiler stuff before, so I *could* do that too.
However, my time frame is more like "next few weeks".
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
9:05 p.m.
New subject: [AC 21.5] -Wdeclaration-after-statement support
APPROVE COMMIT 21.5
In spite of the fact that this currently only affects lib-src, I am
committing this patch anyway. I was waiting until I could rework the
warning flag determination logic, but I decided that that is
conceptually distinct from this patch.
--
Jerry James, Assistant Professor james(a)xemacs.org
Computer Science Department http://www.cs.usu.edu/~jerry/
Utah State University
6742
days inactive
6759
days old
6 comments
2 participants
participants (2)
-
Jerry James -
Stephen J. Turnbull