Mercurial > hg > xemacs-beta
diff configure.ac @ 4932:8b63e21b0436
fix compile issues with gcc 4
-------------------- ChangeLog entries follow: --------------------
ChangeLog addition:
2010-01-24 Ben Wing <ben@xemacs.org>
* aclocal.m4 (XE_SHLIB_STUFF):
Use -export-all-symbols instead of -export-dynamic on PE targets
(Cygwin and MinGW).
* configure.ac (XE_EXPAND_VARIABLE):
* configure.ac (TAB):
Create variable XEMACS_CC_GPP to check whether we're running g++.
Don't just check for an executable called `g++' -- it might be
called g++-4 or whatever. Instead, check for either named `g++*'
or claiming to be g++ when called with --version. Rewrite code do
use the variable.
Add -fno-strict-aliasing to optimization flags when GCC and
optimized, and in all cases with g++, since under these circumstances
strict aliasing is otherwise assumed, and XEmacs can't easily be
made to respect its restrictions.
* configure: Regenerate.
lib-src/ChangeLog addition:
2010-01-24 Ben Wing <ben@xemacs.org>
* fakemail.c (args_size):
* fakemail.c (parse_header):
* ootags.c (C_entries):
Fix warnings about possible use of uninitialized vars.
lwlib/ChangeLog addition:
2010-01-24 Ben Wing <ben@xemacs.org>
* xlwgauge.c (GaugeResize):
* xlwgauge.c (GaugeSize):
Fix warnings about possible use of uninitialized vars.
modules/ChangeLog addition:
2010-01-24 Ben Wing <ben@xemacs.org>
* postgresql/postgresql.c (CHECK_LIVE_CONNECTION):
* postgresql/postgresql.c (print_pgconn):
* postgresql/postgresql.c (Fpq_connectdb):
* postgresql/postgresql.c (Fpq_connect_start):
* postgresql/postgresql.c (Fpq_exec):
* postgresql/postgresql.c (Fpq_get_result):
Fix g++ 4.3 complaints about implicit conversions of string
literals (const char *) to char *.
src/ChangeLog addition:
2010-01-24 Ben Wing <ben@xemacs.org>
* chartab.c (decode_char_table_range):
* extents.c (extent_fragment_update):
* objects-msw.c (initialize_font_instance):
* process.c (Fgetenv):
* redisplay-output.c (get_next_display_block):
Fix warnings about possible use of uninitialized vars.
* compiler.h:
* compiler.h (REGISTER):
* event-stream.c (is_scrollbar_event):
* window.c (window_scrollbar_width):
* window.c (window_scrollbar_height):
* window.c (window_left_window_gutter_width):
* window.c (window_right_window_gutter_width):
Add USED_IF_SCROLLBARS. Use it to fix warnings about unused
vars when --with-scrollbars=no.
* config.h.in:
Change comment to explain better why DECLARE_INLINE_HEADER
is needed.
* dialog-msw.c:
* emacs.c (SHEBANG_EXE_PROGNAME_LENGTH):
* emacs.c (main_1):
* event-msw.c (struct mswin_message_debug):
* event-msw.c (debug_output_mswin_message):
* font-mgr.c:
* font-mgr.c (Ffc_config_filename):
* glyphs-msw.c (struct):
* glyphs-msw.c (bitmap_table):
* glyphs-x.c (update_widget_face):
* intl-win32.c (struct lang_to_string):
* intl-win32.c (lang_to_string_table):
* nas.c:
* objects-xlike-inc.c:
* objects-xlike-inc.c (xft_find_charset_font):
* syswindows.h:
* win32.c (mswindows_output_last_error):
Fix g++ 4.3 complaints about implicit conversions of string
literals (const char *) to char *.
* lisp.h:
G++ 4.3 needs #include <limits> to avoid errors about min/max.
* lisp.h (disabled_assert_with_message):
Use disabled_assert* whenever asserts are disabled. Rewrite
disabled_assert* to avoid complaints about unused vars by
pretending to use the vars but casting them to (void).
Remove code that defined assert() weirdly if DEBUG_XEMACS but
not USE_ASSERTIONS -- configure sets USE_ASSERTIONS automatically
when DEBUG_XEMACS, and if the user has forced it off, then
so be it.
* lisp.h (SYMBOL_KEYWORD):
Put some of the combined `extern Lisp_Object's back under
the file they are declared in. Cosmetic fix.
* number.h:
Remove `extern Lisp_Object' decls that duplicate lisp.h,
since they have different C vs. C++ linkage.
| author | Ben Wing <ben@xemacs.org> |
|---|---|
| date | Sun, 24 Jan 2010 22:04:58 -0600 |
| parents | bde90bc762f2 |
| children | 349f01075eb7 |
line wrap: on
line diff
--- a/configure.ac Thu Jan 21 00:46:37 2010 -0600 +++ b/configure.ac Sun Jan 24 22:04:58 2010 -0600 @@ -1591,6 +1591,31 @@ test -n "$with_xemacs_compiler" && XEMACS_CC="$with_xemacs_compiler" : ${XEMACS_CC:="$CC"} +dnl Are we using g++? + +AC_MSG_CHECKING(whether we are using g++) + +dnl Is it named g++*? +XEMACS_CC_GPP=no +case "$XEMACS_CC" in + *g++* ) XEMACS_CC_GPP=yes ;; +esac + +dnl If it's known to be in the GCC family, does it claim to be g++? We don't +dnl run this test unless we know it's GCC-like, because not all compilers +dnl accept --version. +dnl +dnl It's not clear we need both tests, but it doesn't hurt. g++ might not +dnl have always output "g++" as the first thing in its version string, and +dnl might not always in the future. +if test "$XEMACS_CC_GPP" = "no" -a "$GCC" = "yes"; then + case "`$XEMACS_CC --version`" in + g++* ) XEMACS_CC_GPP=yes ;; + esac +fi + +AC_MSG_RESULT($XEMACS_CC_GPP) + dnl Figure out what C preprocessor to use. dnl On Sun systems, people sometimes set up the variable CPP @@ -1877,24 +1902,13 @@ xemacs_cc_cc_mismatch=no if test "$CC" != "$XEMACS_CC"; then - case "$XEMACS_CC" in - *g++* ) - if test "$GCC" != "yes"; then - AC_MSG_WARN([CC and g++ are mismatched; XE_CFLAGS may be wrong]) - xemacs_cc_cc_mismatch=yes - fi - ;; - esac - if test -n "$GCC"; then - case $XEMACS_CC in - *g++* ) - # it's as expected, do nothing - ;; - * ) - AC_MSG_WARN([gcc and XEMACS_CC are mismatched; XE_CFLAGS may be wrong]) - xemacs_cc_cc_mismatch=yes - ;; - esac + if test "$XEMACS_CC_GPP" = "yes" -a "$GCC" != "yes"; then + AC_MSG_WARN([CC and g++ are mismatched; XE_CFLAGS may be wrong]) + xemacs_cc_cc_mismatch=yes + fi + if test -n "$GCC" -a "$XEMACS_CC_GPP" != "yes" ; then + AC_MSG_WARN([gcc and XEMACS_CC are mismatched; XE_CFLAGS may be wrong]) + xemacs_cc_cc_mismatch=yes fi dnl #### Add other detectable mismatches here. fi @@ -1919,12 +1933,17 @@ fi AC_MSG_CHECKING([for preferred optimization flags]) -if test "$with_optimization" = "yes" ; then - if test "$cflags_optimization_specified" = "no"; then +if test "$cflags_optimization_specified" = "no"; then + if test "$with_optimization" = "yes" ; then dnl Following values of cflags_optimization are known to work well. dnl Should we take debugging options into consideration? if test "$GCC" = "yes"; then - with_cflags_optimization="-O3" + dnl If you want to remove the -fno-strict-aliasing, then you will have + dnl to rewrite all cases of "type-punning" expressions like + dnl (* (foo *) (&bar)), which occur all over XEmacs, by making `bar' a + dnl union (use an anonymous union if you're willing to put in an ifdef + dnl for anon-union support, falling back to the type-punned expression. + with_cflags_optimization="-O3 -fno-strict-aliasing" elif test "$__SUNPRO_C" = "yes"; then case "$opsys" in sol2 ) with_cflags_optimization="-xO4" ;; @@ -1940,9 +1959,12 @@ else with_cflags_optimization="-O" ;dnl The only POSIX-approved flag fi + elif test "$XEMACS_CC_GPP" = "yes" ; then + dnl Fuck me! g++ v4 turns on strict aliasing by default, even without + dnl optimization. See comment above about why we can't have strict + dnl aliasing currently in XEmacs. + with_cflags_optimization="-fno-strict-aliasing" fi -else - with_cflags_optimization= fi AC_MSG_RESULT([${with_cflags_optimization}]) @@ -2005,7 +2027,7 @@ dnl But gcc warns about -Weffc++ in C compiles. dnl With g++, there is no effective way to use -Wunused-parameter without dnl some very ugly code changes. - if test "$with_xemacs_compiler" = "g++"; then + if test "$XEMACS_CC_GPP" = "yes"; then xe_cflags_warning="$with_cflags_warning -Weffc++" elif test "$__GCC" -ge 3; then xe_cflags_warning="$with_cflags_warning -Wunused-parameter" @@ -3312,7 +3334,7 @@ dnl If g++ is used, we have to explicitly link modules with -lstdc++ on Cygwin dnl to avoid undefined symbol errors. This will never hurt, so just do it on dnl all platforms in case others have the same brain damage. -if test "$with_xemacs_compiler" = "g++"; then +if test "$XEMACS_CC_GPP" = "yes"; then LIBSTDCPP=-lstdc++ else LIBSTDCPP= @@ -5490,24 +5512,24 @@ elif test "$with_debug_malloc" = "yes"; then AC_DEFINE(USE_DEBUG_MALLOC) AC_DEFINE(USE_SYSTEM_MALLOC) fi -test "$GCC" = "yes" && AC_DEFINE(USE_GCC) -test "$with_xemacs_compiler" = "g++" && AC_DEFINE(USE_GPLUSPLUS) +test "$GCC" = "yes" && AC_DEFINE(USE_GCC) +test "$XEMACS_CC_GPP" = "yes" && AC_DEFINE(USE_GPLUSPLUS) test "$with_external_widget" = "yes" && AC_DEFINE(EXTERNAL_WIDGET) -test "$with_kkcc" = "yes" && AC_DEFINE(USE_KKCC) -test "$with_newgc" = "yes" && AC_DEFINE(NEW_GC) -test "$have_vdb_posix" = "yes" && AC_DEFINE(VDB_POSIX) -test "$have_vdb_fake" = "yes" && AC_DEFINE(VDB_FAKE) -test "$with_quick_build" = "yes" && AC_DEFINE(QUICK_BUILD) -test "$with_purify" = "yes" && AC_DEFINE(PURIFY) -test "$with_quantify" = "yes" && AC_DEFINE(QUANTIFY) -test "$with_valgrind" = "yes" && AC_DEFINE(USE_VALGRIND) -test "$with_pop" = "yes" && AC_DEFINE(MAIL_USE_POP) -test "$with_kerberos" = "yes" && AC_DEFINE(KERBEROS) -test "$with_hesiod" = "yes" && AC_DEFINE(HESIOD) -test "$with_union_type" = "yes" && AC_DEFINE(USE_UNION_TYPE) -test "$with_pdump" = "yes" && AC_DEFINE(PDUMP) -test "$with_dump_in_exec" = "yes" && AC_DEFINE(DUMP_IN_EXEC) -test "$with_ipv6_cname" = "yes" && AC_DEFINE(IPV6_CANONICALIZE) +test "$with_kkcc" = "yes" && AC_DEFINE(USE_KKCC) +test "$with_newgc" = "yes" && AC_DEFINE(NEW_GC) +test "$have_vdb_posix" = "yes" && AC_DEFINE(VDB_POSIX) +test "$have_vdb_fake" = "yes" && AC_DEFINE(VDB_FAKE) +test "$with_quick_build" = "yes" && AC_DEFINE(QUICK_BUILD) +test "$with_purify" = "yes" && AC_DEFINE(PURIFY) +test "$with_quantify" = "yes" && AC_DEFINE(QUANTIFY) +test "$with_valgrind" = "yes" && AC_DEFINE(USE_VALGRIND) +test "$with_pop" = "yes" && AC_DEFINE(MAIL_USE_POP) +test "$with_kerberos" = "yes" && AC_DEFINE(KERBEROS) +test "$with_hesiod" = "yes" && AC_DEFINE(HESIOD) +test "$with_union_type" = "yes" && AC_DEFINE(USE_UNION_TYPE) +test "$with_pdump" = "yes" && AC_DEFINE(PDUMP) +test "$with_dump_in_exec" = "yes" && AC_DEFINE(DUMP_IN_EXEC) +test "$with_ipv6_cname" = "yes" && AC_DEFINE(IPV6_CANONICALIZE)