newlib_cv_ vs libc_cv_ autoconf cache vars prefix

newlib_cv_ vs libc_cv_ autoconf cache vars prefix

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 458 bytes --]

the newlib/libgloss configure scripts use both libc_cv_ and newlib_cv_ as
their autoconf cache var prefixes.  this really should be just one name.
i'm inclined to use newlib_cv_ rather than the more generic libc_cv_.  any
preferences or reasons to pick one or the other ?

libgloss also has cache vars.  i'm inclined to have it use newlib_cv_ too
since, as a project, it's carved out the namespace, and doesn't really need
its own libgloss_cv_ prefix.
-mike

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: newlib_cv_ vs libc_cv_ autoconf cache vars prefix

[Corinna Vinschen]

On Jan 28 04:44, Mike Frysinger wrote:
> the newlib/libgloss configure scripts use both libc_cv_ and newlib_cv_ as
> their autoconf cache var prefixes.  this really should be just one name.
> i'm inclined to use newlib_cv_ rather than the more generic libc_cv_.  any
> preferences or reasons to pick one or the other ?
> 
> libgloss also has cache vars.  i'm inclined to have it use newlib_cv_ too
> since, as a project, it's carved out the namespace, and doesn't really need
> its own libgloss_cv_ prefix.
> -mike

AFAICS, only six libc_cv_* and two newlib_cv_* vars exist at all, and only
two of the libc_cv_* vars are AC_SUBST'ed and show up in the Makefile.
Using the same prefix for all vars, it would be the smaller change to
switch the 2 newlib_cv's to libc_cv.  The result would be the same.


Corinna

Re: newlib_cv_ vs libc_cv_ autoconf cache vars prefix

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 1152 bytes --]

On 31 Jan 2022 15:17, Corinna Vinschen wrote:
> On Jan 28 04:44, Mike Frysinger wrote:
> > the newlib/libgloss configure scripts use both libc_cv_ and newlib_cv_ as
> > their autoconf cache var prefixes.  this really should be just one name.
> > i'm inclined to use newlib_cv_ rather than the more generic libc_cv_.  any
> > preferences or reasons to pick one or the other ?
> > 
> > libgloss also has cache vars.  i'm inclined to have it use newlib_cv_ too
> > since, as a project, it's carved out the namespace, and doesn't really need
> > its own libgloss_cv_ prefix.
> 
> AFAICS, only six libc_cv_* and two newlib_cv_* vars exist at all, and only
> two of the libc_cv_* vars are AC_SUBST'ed and show up in the Makefile.
> Using the same prefix for all vars, it would be the smaller change to
> switch the 2 newlib_cv's to libc_cv.  The result would be the same.

the reason i'm not a fan of libc_cv is that glibc uses that namespace.  if you
want to setup a config.site cache for a target that builds glibc & newlib, then
it's hard to avoid namespace conflicts.  but if we use newlib_cv_, then there's
no such collision.
-mike

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: newlib_cv_ vs libc_cv_ autoconf cache vars prefix

[Corinna Vinschen]

On Jan 31 22:12, Mike Frysinger wrote:
> On 31 Jan 2022 15:17, Corinna Vinschen wrote:
> > On Jan 28 04:44, Mike Frysinger wrote:
> > > the newlib/libgloss configure scripts use both libc_cv_ and newlib_cv_ as
> > > their autoconf cache var prefixes.  this really should be just one name.
> > > i'm inclined to use newlib_cv_ rather than the more generic libc_cv_.  any
> > > preferences or reasons to pick one or the other ?
> > > 
> > > libgloss also has cache vars.  i'm inclined to have it use newlib_cv_ too
> > > since, as a project, it's carved out the namespace, and doesn't really need
> > > its own libgloss_cv_ prefix.
> > 
> > AFAICS, only six libc_cv_* and two newlib_cv_* vars exist at all, and only
> > two of the libc_cv_* vars are AC_SUBST'ed and show up in the Makefile.
> > Using the same prefix for all vars, it would be the smaller change to
> > switch the 2 newlib_cv's to libc_cv.  The result would be the same.
> 
> the reason i'm not a fan of libc_cv is that glibc uses that namespace.  if you
> want to setup a config.site cache for a target that builds glibc & newlib, then
> it's hard to avoid namespace conflicts.  but if we use newlib_cv_, then there's
> no such collision.
> -mike

Good point.  Feel free to make the change.


Corinna

[PATCH 1/2] newlib: drop unused cache vars from Makefiles

[Mike Frysinger]

These aren't used in any of the makefiles, so there's no point in
exporting these.  These are only checked in the configure script.
---
 newlib/Makefile.in  | 2 --
 newlib/configure    | 8 ++------
 newlib/configure.ac | 2 --
 3 files changed, 2 insertions(+), 10 deletions(-)

diff --git a/newlib/Makefile.in b/newlib/Makefile.in
index d2a0d4422f9a..5d49279ce48b 100644
--- a/newlib/Makefile.in
+++ b/newlib/Makefile.in
@@ -407,8 +407,6 @@ htmldir = @htmldir@
 includedir = @includedir@
 infodir = @infodir@
 install_sh = @install_sh@
-libc_cv_cc_loop_to_function = @libc_cv_cc_loop_to_function@
-libc_cv_initfinit_array = @libc_cv_initfinit_array@
 libdir = @libdir@
 libexecdir = @libexecdir@
 libm_machine_dir = @libm_machine_dir@
diff --git a/newlib/configure b/newlib/configure
index 65b93630315e..ca8953933cda 100755
--- a/newlib/configure
+++ b/newlib/configure
@@ -636,8 +636,6 @@ LTLIBOBJS
 LIBOBJS
 HAVE_LONG_DOUBLE_FALSE
 HAVE_LONG_DOUBLE_TRUE
-libc_cv_cc_loop_to_function
-libc_cv_initfinit_array
 NEWLIB_PATCHLEVEL_VERSION
 NEWLIB_MINOR_VERSION
 NEWLIB_MAJOR_VERSION
@@ -12008,7 +12006,7 @@ else
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<_LT_EOF
-#line 12011 "configure"
+#line 12009 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -12114,7 +12112,7 @@ else
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<_LT_EOF
-#line 12117 "configure"
+#line 12115 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -13300,7 +13298,6 @@ rm -f conftest*
 fi
 { $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_initfinit_array" >&5
 $as_echo "$libc_cv_initfinit_array" >&6; }
-
 if test $libc_cv_initfinit_array = yes; then
 
 $as_echo "#define _HAVE_INITFINI_ARRAY 1" >>confdefs.h
@@ -13341,7 +13338,6 @@ $as_echo "#define _HAVE_CC_INHIBIT_LOOP_TO_LIBCALL 1" >>confdefs.h
 fi
 
 
-
   { $as_echo "$as_me:${as_lineno-$LINENO}: checking for long double" >&5
 $as_echo_n "checking for long double... " >&6; }
 if ${ac_cv_type_long_double+:} false; then :
diff --git a/newlib/configure.ac b/newlib/configure.ac
index c164e5a6494f..ce36ea19ac10 100644
--- a/newlib/configure.ac
+++ b/newlib/configure.ac
@@ -656,7 +656,6 @@ else
   libc_cv_initfinit_array=no
 fi
 rm -f conftest*])
-AC_SUBST(libc_cv_initfinit_array)
 if test $libc_cv_initfinit_array = yes; then
   AC_DEFINE(_HAVE_INITFINI_ARRAY, 1, [Define if the linker supports .preinit_array/.init_array/.fini_array sections.])
 fi
@@ -678,7 +677,6 @@ rm -f conftest*])
 if test $libc_cv_cc_loop_to_function = yes; then
   AC_DEFINE(_HAVE_CC_INHIBIT_LOOP_TO_LIBCALL, 1, [Define if compiler supports -fno-tree-loop-distribute-patterns.])
 fi
-AC_SUBST(libc_cv_cc_loop_to_function)
 
 AC_TYPE_LONG_DOUBLE
 dnl Export the setting for our installed headers to check.
-- 
2.34.1

[PATCH 2/2] newlib: rename libc_cv_ prefix to newlib_cv_

[Mike Frysinger]

We've been using both libc_cv_ and newlib_cv_ for our cache vars.
Let's consolidate on newlib_cv_ to avoid conflicts with glibc which
is already using the libc_cv_ prefix.
---
 newlib/configure      | 26 +++++++++++++-------------
 newlib/configure.ac   | 18 +++++++++---------
 newlib/configure.host |  2 +-
 3 files changed, 23 insertions(+), 23 deletions(-)

diff --git a/newlib/configure b/newlib/configure
index ca8953933cda..a39a3e51afcd 100755
--- a/newlib/configure
+++ b/newlib/configure
@@ -13269,7 +13269,7 @@ $as_echo "#define _ICONV_FROM_ENCODING_WIN_1258 1" >>confdefs.h
 
 { $as_echo "$as_me:${as_lineno-$LINENO}: checking for .preinit_array/.init_array/.fini_array support" >&5
 $as_echo_n "checking for .preinit_array/.init_array/.fini_array support... " >&6; }
-if ${libc_cv_initfinit_array+:} false; then :
+if ${newlib_cv_initfinit_array+:} false; then :
   $as_echo_n "(cached) " >&6
 else
   cat > conftest.c <<EOF
@@ -13287,18 +13287,18 @@ if { ac_try='${CC} $CFLAGS $CPPFLAGS -c -o conftest conftest.c
   test $ac_status = 0; }; }
 then
   if ${READELF} -S conftest | grep -e INIT_ARRAY > /dev/null; then
-    libc_cv_initfinit_array=yes
+    newlib_cv_initfinit_array=yes
   else
-    libc_cv_initfinit_array=no
+    newlib_cv_initfinit_array=no
   fi
 else
-  libc_cv_initfinit_array=no
+  newlib_cv_initfinit_array=no
 fi
 rm -f conftest*
 fi
-{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_initfinit_array" >&5
-$as_echo "$libc_cv_initfinit_array" >&6; }
-if test $libc_cv_initfinit_array = yes; then
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $newlib_cv_initfinit_array" >&5
+$as_echo "$newlib_cv_initfinit_array" >&6; }
+if test $newlib_cv_initfinit_array = yes; then
 
 $as_echo "#define _HAVE_INITFINI_ARRAY 1" >>confdefs.h
 
@@ -13308,7 +13308,7 @@ fi
 __attribute__ ((__optimize__))" >&5
 $as_echo_n "checking if $CC accepts -fno-tree-loop-distribute-patterns with \
 __attribute__ ((__optimize__))... " >&6; }
-if ${libc_cv_cc_loop_to_function+:} false; then :
+if ${newlib_cv_cc_loop_to_function+:} false; then :
   $as_echo_n "(cached) " >&6
 else
   cat > conftest.c <<EOF
@@ -13316,7 +13316,7 @@ void
 __attribute__ ((__optimize__ ("-fno-tree-loop-distribute-patterns")))
 foo (void) {}
 EOF
-libc_cv_cc_loop_to_function=no
+newlib_cv_cc_loop_to_function=no
 if { ac_try='${CC-cc} $CFLAGS $CPPFLAGS -c
 			    -fno-tree-loop-distribute-patterns conftest.c'
   { { eval echo "\"\$as_me\":${as_lineno-$LINENO}: \"$ac_try\""; } >&5
@@ -13325,13 +13325,13 @@ if { ac_try='${CC-cc} $CFLAGS $CPPFLAGS -c
   $as_echo "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5
   test $ac_status = 0; }; }
 then
-  libc_cv_cc_loop_to_function=yes
+  newlib_cv_cc_loop_to_function=yes
 fi
 rm -f conftest*
 fi
-{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_cc_loop_to_function" >&5
-$as_echo "$libc_cv_cc_loop_to_function" >&6; }
-if test $libc_cv_cc_loop_to_function = yes; then
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $newlib_cv_cc_loop_to_function" >&5
+$as_echo "$newlib_cv_cc_loop_to_function" >&6; }
+if test $newlib_cv_cc_loop_to_function = yes; then
 
 $as_echo "#define _HAVE_CC_INHIBIT_LOOP_TO_LIBCALL 1" >>confdefs.h
 
diff --git a/newlib/configure.ac b/newlib/configure.ac
index ce36ea19ac10..2e6a80a5f8b3 100644
--- a/newlib/configure.ac
+++ b/newlib/configure.ac
@@ -637,7 +637,7 @@ fi;
 NEWLIB_ICONV_DEFINES
 
 AC_CACHE_CHECK(for .preinit_array/.init_array/.fini_array support,
-	       libc_cv_initfinit_array, [dnl
+	       newlib_cv_initfinit_array, [dnl
 cat > conftest.c <<EOF
 int _start (void) { return 0; }
 int __start (void) { return 0; }
@@ -648,33 +648,33 @@ if AC_TRY_COMMAND([${CC} $CFLAGS $CPPFLAGS -c -o conftest conftest.c
 							1>&AS_MESSAGE_LOG_FD])
 then
   if ${READELF} -S conftest | grep -e INIT_ARRAY > /dev/null; then
-    libc_cv_initfinit_array=yes
+    newlib_cv_initfinit_array=yes
   else
-    libc_cv_initfinit_array=no
+    newlib_cv_initfinit_array=no
   fi
 else
-  libc_cv_initfinit_array=no
+  newlib_cv_initfinit_array=no
 fi
 rm -f conftest*])
-if test $libc_cv_initfinit_array = yes; then
+if test $newlib_cv_initfinit_array = yes; then
   AC_DEFINE(_HAVE_INITFINI_ARRAY, 1, [Define if the linker supports .preinit_array/.init_array/.fini_array sections.])
 fi
 
 AC_CACHE_CHECK(if $CC accepts -fno-tree-loop-distribute-patterns with \
-__attribute__ ((__optimize__)), libc_cv_cc_loop_to_function, [dnl
+__attribute__ ((__optimize__)), newlib_cv_cc_loop_to_function, [dnl
 cat > conftest.c <<EOF
 void
 __attribute__ ((__optimize__ ("-fno-tree-loop-distribute-patterns")))
 foo (void) {}
 EOF
-libc_cv_cc_loop_to_function=no
+newlib_cv_cc_loop_to_function=no
 if AC_TRY_COMMAND([${CC-cc} $CFLAGS $CPPFLAGS -c
 			    -fno-tree-loop-distribute-patterns conftest.c])
 then
-  libc_cv_cc_loop_to_function=yes
+  newlib_cv_cc_loop_to_function=yes
 fi
 rm -f conftest*])
-if test $libc_cv_cc_loop_to_function = yes; then
+if test $newlib_cv_cc_loop_to_function = yes; then
   AC_DEFINE(_HAVE_CC_INHIBIT_LOOP_TO_LIBCALL, 1, [Define if compiler supports -fno-tree-loop-distribute-patterns.])
 fi
 
diff --git a/newlib/configure.host b/newlib/configure.host
index bf4a555e6da7..893a698ed6d9 100644
--- a/newlib/configure.host
+++ b/newlib/configure.host
@@ -126,7 +126,7 @@ case "${host_cpu}" in
   amdgcn*)
 	newlib_cflags="${newlib_cflags} -D__DYNAMIC_REENT__"
 	machine_dir=amdgcn
-	libc_cv_initfinit_array=yes
+	newlib_cv_initfinit_array=yes
 	;;
   arc*)
 	machine_dir=arc
-- 
2.34.1

Re: [PATCH 1/2] newlib: drop unused cache vars from Makefiles

[Corinna Vinschen]

On Feb  1 21:52, Mike Frysinger wrote:
> These aren't used in any of the makefiles, so there's no point in
> exporting these.  These are only checked in the configure script.
> ---
>  newlib/Makefile.in  | 2 --
>  newlib/configure    | 8 ++------
>  newlib/configure.ac | 2 --
>  3 files changed, 2 insertions(+), 10 deletions(-)

Cover letter?  Pleeeeeease?

Patchset is ok to push.


Thanks,
Corinna