[PATCH v2 0/4] Avoid relying on autoconf internals in `preconfigure' scripts

[PATCH v2 0/4] Avoid relying on autoconf internals in `preconfigure' scripts

[Maciej W. Rozycki]

Hi,

 I have discovered that we have two kinds of `preconfigure' scripts, 
handwritten ones and autoconf-generated ones.  While it seems perfectly 
fine to have these scripts handwritten, some of them wire into autoconf 
internals or use an ad-hoc way to report errors.  The scripts for the 
C-SKY, m68k, MIPS and RISC-V targets fall into this category.

 This small patch series addresses this problem by converting these 
scripts to autoconf templates with suitable adjustments, then processed 
with autoconf to re-create functionally equivalent almost identical 
scripts.

 The RISC-V script has been verified to still correctly build the library.  
All have been visually inspected to ensure functional equivalence; changes 
are so minimal for this to be straightforward.  There's some noise though 
from a formatting fix to replace octets of spaces with tabs in the new 
templates.

 OK to apply?

  Maciej

[PATCH v2 1/4] C-SKY: Use an autoconf template to produce `preconfigure'

[Maciej W. Rozycki]

Avoid fiddling with autoconf internals and use AC_DEFINE_UNQUOTED to 
define macros in the configuration headers rather than handcoding an 
equivalent shell sequence with the use of the `as_echo' undocumented 
variable.

Switch to using AC_MSG_ERROR rather than `echo' and `exit' directly for 
error handling.  Owing to the lack of any kind of error annotation it 
makes it difficult to spot the message in the flood in a parallel build 
and neither it is logged in `config.log'.
---
New change in v2.
---
 sysdeps/csky/preconfigure    |   27 +++++++++++++--------
 sysdeps/csky/preconfigure.ac |   55 +++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 72 insertions(+), 10 deletions(-)

glibc-csky-preconfigure-as-fn-error.diff
Index: glibc/sysdeps/csky/preconfigure
===================================================================
--- glibc.orig/sysdeps/csky/preconfigure
+++ glibc/sysdeps/csky/preconfigure
@@ -1,3 +1,6 @@
+# This file is generated from configure.ac by Autoconf.  DO NOT EDIT!
+# Local preconfigure fragment for sysdeps/csky
+
 case "$machine" in
 csky*)
     abi=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null |
@@ -11,16 +14,14 @@ csky*)
 
     case "$abi" in
     1)
-        echo "glibc does not support abiv1 yet" >&2
-        exit 1
-        ;;
+	as_fn_error 1 "glibc does not support abiv1 yet" "$LINENO" 5
+	;;
     2)
-        machine=abiv2
-        ;;
+	machine=abiv2
+	;;
     *)
-        echo "Unknown abi" >&2
-        exit 1
-        ;;
+	as_fn_error 1 "Unknown abi" "$LINENO" 5
+	;;
     esac
 
     # __CSKY_HARD_FLOAT_ABI__ was added on gcc 11 to specify whether
@@ -48,7 +49,13 @@ csky*)
     base_machine=csky
     machine=csky/$machine
 
-    $as_echo "#define CSKYABI $abi" >>confdefs.h
-    $as_echo "#define CSKY_HARD_FLOAT $with_fp_cond" >>confdefs.h
+    cat >>confdefs.h <<_ACEOF
+#define CSKYABI $abi
+_ACEOF
+
+    cat >>confdefs.h <<_ACEOF
+#define CSKY_HARD_FLOAT $with_fp_cond
+_ACEOF
+
     ;;
 esac
Index: glibc/sysdeps/csky/preconfigure.ac
===================================================================
--- /dev/null
+++ glibc/sysdeps/csky/preconfigure.ac
@@ -0,0 +1,55 @@
+GLIBC_PROVIDES[]dnl See aclocal.m4 in the top level source directory.
+# Local preconfigure fragment for sysdeps/csky
+
+case "$machine" in
+csky*)
+    abi=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null |
+      sed -n 's/^#define __CSKYABI__ \(.*\)/\1/p'`
+    hard_float=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null |
+      sed -n 's/^#define __CSKY_HARD_FLOAT__ \(.*\)/\1/p'`
+    hard_float_sf=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null |
+      sed -n 's/^#define __CSKY_HARD_FLOAT_FPU_SF__ \(.*\)/\1/p'`
+    hard_float_abi=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null |
+      sed -n 's/^#define __CSKY_HARD_FLOAT_ABI__ \(.*\)/\1/p'`
+
+    case "$abi" in
+    1)
+	AC_MSG_ERROR([glibc does not support abiv1 yet], [1])
+	;;
+    2)
+	machine=abiv2
+	;;
+    *)
+	AC_MSG_ERROR([Unknown abi], [1])
+	;;
+    esac
+
+    # __CSKY_HARD_FLOAT_ABI__ was added on gcc 11 to specify whether
+    # -mfloat-abi=hard is set.  On older gcc, the float ABI is defined solely
+    # with __CSKY_HARD_FLOAT__.  If __CSKY_HARD_FLOAT__ is set, it can be
+    # either a hard-float ABI (gcc older than 11, or gcc11 -mfloat-abi=hard
+    # (__CSKY_HARD_FLOAT_ABI__ is set) or -mfloat-abi=softfp
+    # (__CSKY_HARD_FLOAT_ABI__ is not set). To be compatible with older gcc,
+    # use __CSKY_HARD_FLOAT_FPU_SF__ identify if -mfloat-abi is supported,
+    # because it is added to gcc at the same time as -mfloat-abi.
+    if test -n "$hard_float"; then
+	if test -z "$hard_float_sf"; then
+	    with_fp_cond=1
+	else
+	    if test -n "$hard_float_abi"; then
+		with_fp_cond=1
+	    else
+		with_fp_cond=0
+	    fi
+	fi
+    else
+	with_fp_cond=0
+    fi
+
+    base_machine=csky
+    machine=csky/$machine
+
+    AC_DEFINE_UNQUOTED([CSKYABI], [$abi])
+    AC_DEFINE_UNQUOTED([CSKY_HARD_FLOAT], [$with_fp_cond])
+    ;;
+esac

Re: [PATCH v2 1/4] C-SKY: Use an autoconf template to produce `preconfigure'

[Andreas Schwab]

On Mai 12 2022, Maciej W. Rozycki wrote:

> Avoid fiddling with autoconf internals and use AC_DEFINE_UNQUOTED to 
> define macros in the configuration headers rather than handcoding an 
> equivalent shell sequence with the use of the `as_echo' undocumented 
> variable.
>
> Switch to using AC_MSG_ERROR rather than `echo' and `exit' directly for 
> error handling.  Owing to the lack of any kind of error annotation it 
> makes it difficult to spot the message in the flood in a parallel build 
> and neither it is logged in `config.log'.

Ok.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

[PATCH v2 2/4] m68k: Use an autoconf template to produce `preconfigure'

[Maciej W. Rozycki]

Switch to using AC_MSG_ERROR rather than `echo' and `exit' directly for 
error handling.  Owing to the lack of any kind of error annotation it 
makes it difficult to spot the message in the flood in a parallel build 
and neither it is logged in `config.log'.
---
New change in v2.
---
 sysdeps/m68k/preconfigure    |    6 ++++--
 sysdeps/m68k/preconfigure.ac |   23 +++++++++++++++++++++++
 2 files changed, 27 insertions(+), 2 deletions(-)

glibc-m68k-preconfigure-as-fn-error.diff
Index: glibc/sysdeps/m68k/preconfigure
===================================================================
--- glibc.orig/sysdeps/m68k/preconfigure
+++ glibc/sysdeps/m68k/preconfigure
@@ -1,3 +1,6 @@
+# This file is generated from configure.ac by Autoconf.  DO NOT EDIT!
+# Local preconfigure fragment for sysdeps/m68k
+
 # This fragment canonicalizes the machine names for m68k variants.
 
 case "$machine" in
@@ -10,8 +13,7 @@ m68k)		variant=`(echo "#ifdef __mcoldfir
 			 $CC $CFLAGS $CPPFLAGS -E - |
 			 grep '^[a-z]'`
 		if test -z "$variant"; then
-		  echo >&2 "Cannot determine m68k processor variant"
-		  exit 1
+		  as_fn_error 1 "Cannot determine m68k processor variant" "$LINENO" 5
 		fi
 		base_machine=m68k machine=m68k/$variant
 		if test "$variant" = "coldfire"; then
Index: glibc/sysdeps/m68k/preconfigure.ac
===================================================================
--- /dev/null
+++ glibc/sysdeps/m68k/preconfigure.ac
@@ -0,0 +1,23 @@
+GLIBC_PROVIDES[]dnl See aclocal.m4 in the top level source directory.
+# Local preconfigure fragment for sysdeps/m68k
+
+# This fragment canonicalizes the machine names for m68k variants.
+
+case "$machine" in
+m680?0)		base_machine=m68k machine=m68k/m680x0/$machine ;;
+m68k)		variant=`(echo "#ifdef __mcoldfire__"
+			  echo "coldfire"
+			  echo "#else"
+			  echo "m680x0/m68020"
+			  echo "#endif") |
+			 $CC $CFLAGS $CPPFLAGS -E - |
+			 grep '^[[a-z]]'`
+		if test -z "$variant"; then
+		  AC_MSG_ERROR([Cannot determine m68k processor variant], [1])
+		fi
+		base_machine=m68k machine=m68k/$variant
+		if test "$variant" = "coldfire"; then
+		  with_fp_cond="defined __mcffpu__"
+		fi
+		;;
+esac

Re: [PATCH v2 2/4] m68k: Use an autoconf template to produce `preconfigure'

[Andreas Schwab]

On Mai 12 2022, Maciej W. Rozycki wrote:

> Switch to using AC_MSG_ERROR rather than `echo' and `exit' directly for 
> error handling.  Owing to the lack of any kind of error annotation it 
> makes it difficult to spot the message in the flood in a parallel build 
> and neither it is logged in `config.log'.

Ok.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

[PATCH v2 3/4] MIPS: Use an autoconf template to produce `preconfigure'

[Maciej W. Rozycki]

Avoid fiddling with autoconf internals and use AC_DEFINE_UNQUOTED to 
define macros in the configuration headers rather than handcoding an 
equivalent shell sequence with the use of the `as_echo' undocumented 
variable.

Similarly use AC_MSG_ERROR for error handling rather than the internal 
undocumented `as_fn_error' variable.  Switch to using 1 as the exit code 
as it makes no sense to refer $? in the contexts involved, it's not a 
command failure handled there.
---
New change in v2.
---
 sysdeps/mips/preconfigure    |    7 +++++--
 sysdeps/mips/preconfigure.ac |   34 ++++++++++++++++++++++++++++++++++
 2 files changed, 39 insertions(+), 2 deletions(-)

glibc-mips-preconfigure-as-fn-error.diff
Index: glibc/sysdeps/mips/preconfigure
===================================================================
--- glibc.orig/sysdeps/mips/preconfigure
+++ glibc/sysdeps/mips/preconfigure
@@ -1,3 +1,6 @@
+# This file is generated from configure.ac by Autoconf.  DO NOT EDIT!
+# Local preconfigure fragment for sysdeps/mips
+
 case "$machine" in
 mips*)
 	abiflag=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define _MIPS_SIM \(.*\)/\1/p'`
@@ -15,7 +18,7 @@ mips*)
 	elif test "$abiflag" = "_ABI64" ; then
 		machine=mips/mips64/n64
 	else
-		as_fn_error $? "Unable to determine ABI." "$LINENO" 5
+		as_fn_error 1 "Unable to determine ABI." "$LINENO" 5
 	fi
 	# $config_machine is not really needed here but the slash after
 	# $machine is needed by the case statement in
@@ -23,7 +26,7 @@ mips*)
 	machine=$machine/$config_machine
 
 	if test "$abiflag" != "_ABIO32" -a "$mips16flag" = "1"; then
-		as_fn_error $? "MIPS16 is only supported with the o32 ABI." "$LINENO" 5
+		as_fn_error 1 "MIPS16 is only supported with the o32 ABI." "$LINENO" 5
 	fi
 
 	with_fp_cond="defined __mips_hard_float"
Index: glibc/sysdeps/mips/preconfigure.ac
===================================================================
--- /dev/null
+++ glibc/sysdeps/mips/preconfigure.ac
@@ -0,0 +1,34 @@
+GLIBC_PROVIDES[]dnl See aclocal.m4 in the top level source directory.
+# Local preconfigure fragment for sysdeps/mips
+
+case "$machine" in
+mips*)
+	abiflag=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define _MIPS_SIM \(.*\)/\1/p'`
+	mips16flag=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __mips16 \(.*\)/\1/p'`
+
+	base_machine=mips
+	if test "$abiflag" = "_ABIO32" ; then
+		if test "$mips16flag" = "1" ; then
+			machine=mips/mips32/mips16
+		else
+			machine=mips/mips32
+		fi
+	elif test "$abiflag" = "_ABIN32" ; then
+		machine=mips/mips64/n32
+	elif test "$abiflag" = "_ABI64" ; then
+		machine=mips/mips64/n64
+	else
+		AC_MSG_ERROR([Unable to determine ABI.], [1])
+	fi
+	# $config_machine is not really needed here but the slash after
+	# $machine is needed by the case statement in
+	# sysdeps/unix/sysv/linux/mips/configure.ac.
+	machine=$machine/$config_machine
+
+	if test "$abiflag" != "_ABIO32" -a "$mips16flag" = "1"; then
+		AC_MSG_ERROR([MIPS16 is only supported with the o32 ABI.], [1])
+	fi
+
+	with_fp_cond="defined __mips_hard_float"
+	;;
+esac

Re: [PATCH v2 3/4] MIPS: Use an autoconf template to produce `preconfigure'

[Andreas Schwab]

On Mai 12 2022, Maciej W. Rozycki wrote:

> Avoid fiddling with autoconf internals and use AC_DEFINE_UNQUOTED to 
> define macros in the configuration headers rather than handcoding an 
> equivalent shell sequence with the use of the `as_echo' undocumented 
> variable.
>
> Similarly use AC_MSG_ERROR for error handling rather than the internal 
> undocumented `as_fn_error' variable.  Switch to using 1 as the exit code 
> as it makes no sense to refer $? in the contexts involved, it's not a 
> command failure handled there.

Ok.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: [PATCH v2 3/4] MIPS: Use an autoconf template to produce `preconfigure'

[Maciej W. Rozycki]

On Fri, 13 May 2022, Andreas Schwab wrote:

> > Similarly use AC_MSG_ERROR for error handling rather than the internal 
> > undocumented `as_fn_error' variable.  Switch to using 1 as the exit code 
> > as it makes no sense to refer $? in the contexts involved, it's not a 
> > command failure handled there.
> 
> Ok.

 Series committed, thank you both for the review.

  Maciej

[PATCH v2 4/4] RISC-V: Use an autoconf template to produce `preconfigure'

[Maciej W. Rozycki]

Avoid fiddling with autoconf internals and use AC_DEFINE_UNQUOTED to 
define macros in the configuration headers rather than handcoding an 
equivalent shell sequence with the use of the `as_echo' undocumented 
variable.

Switch to using AC_MSG_ERROR rather than `echo' and `exit' directly for 
error handling.  Owing to the lack of any kind of error annotation it 
makes it difficult to spot the message in the flood in a parallel build 
and neither it is logged in `config.log'.
---
Changes from v1:

- Convert the script to an autoconf template and use AC_MSG_ERROR.

- Also handle configuration macro definitions, using AC_DEFINE_UNQUOTED.
---
 sysdeps/riscv/preconfigure    |   33 +++++++++++----------
 sysdeps/riscv/preconfigure.ac |   64 ++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 82 insertions(+), 15 deletions(-)

glibc-riscv-preconfigure-as-fn-error.diff
Index: glibc/sysdeps/riscv/preconfigure
===================================================================
--- glibc.orig/sysdeps/riscv/preconfigure
+++ glibc/sysdeps/riscv/preconfigure
@@ -1,3 +1,6 @@
+# This file is generated from configure.ac by Autoconf.  DO NOT EDIT!
+# Local preconfigure fragment for sysdeps/riscv
+
 case "$machine" in
 riscv*)
     xlen=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __riscv_xlen \(.*\)/\1/p'`
@@ -9,8 +12,7 @@ riscv*)
     64 | 32)
 	;;
     *)
-	echo "Unable to determine XLEN" >&2
-	exit 1
+	as_fn_error 1 "Unable to determine XLEN" "$LINENO" 5
 	;;
     esac
 
@@ -20,15 +22,13 @@ riscv*)
 	with_fp_cond=1
 	;;
     32)
-	echo "glibc does not yet support systems with the F but not D extensions" >&2
-	exit 1
+	as_fn_error 1 "glibc does not yet support systems with the F but not D extensions" "$LINENO" 5
 	;;
     "")
 	with_fp_cond=0
 	;;
     *)
-	echo "Unable to determine FLEN" >&2
-	exit 1
+	as_fn_error 1 "Unable to determine FLEN" "$LINENO" 5
 	;;
     esac
 
@@ -37,31 +37,34 @@ riscv*)
 	abi_flen=0
 	;;
     single)
-	echo "glibc does not yet support the single floating-point ABI" >&2
-	exit 1
+	as_fn_error 1 "glibc does not yet support the single floating-point ABI" "$LINENO" 5
 	;;
     double)
 	abi_flen=64
 	;;
     *)
-	echo "Unable to determine floating-point ABI" >&2
-	exit 1
+	as_fn_error 1 "Unable to determine floating-point ABI" "$LINENO" 5
 	;;
     esac
 
     case "$atomic" in
     __riscv_atomic)
-        ;;
+	;;
     *)
-        echo "glibc requires the A extension" >&2
-	exit 1
+	as_fn_error 1 "glibc requires the A extension" "$LINENO" 5
 	;;
     esac
 
     base_machine=riscv
     machine=riscv/rv$xlen/$float_machine
 
-    $as_echo "#define RISCV_ABI_XLEN $xlen" >>confdefs.h
-    $as_echo "#define RISCV_ABI_FLEN $abi_flen" >>confdefs.h
+    cat >>confdefs.h <<_ACEOF
+#define RISCV_ABI_XLEN $xlen
+_ACEOF
+
+    cat >>confdefs.h <<_ACEOF
+#define RISCV_ABI_FLEN $abi_flen
+_ACEOF
+
     ;;
 esac
Index: glibc/sysdeps/riscv/preconfigure.ac
===================================================================
--- /dev/null
+++ glibc/sysdeps/riscv/preconfigure.ac
@@ -0,0 +1,64 @@
+GLIBC_PROVIDES[]dnl See aclocal.m4 in the top level source directory.
+# Local preconfigure fragment for sysdeps/riscv
+
+case "$machine" in
+riscv*)
+    xlen=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __riscv_xlen \(.*\)/\1/p'`
+    flen=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __riscv_flen \(.*\)/\1/p'`
+    float_abi=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __riscv_float_abi_\([^ ]*\) .*/\1/p'`
+    atomic=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | grep '#define __riscv_atomic' | cut -d' ' -f2`
+
+    case "$xlen" in
+    64 | 32)
+	;;
+    *)
+	AC_MSG_ERROR([Unable to determine XLEN], [1])
+	;;
+    esac
+
+    case "$flen" in
+    64)
+	float_machine=rvd
+	with_fp_cond=1
+	;;
+    32)
+	AC_MSG_ERROR([glibc does not yet support systems with the F but not D extensions], 1)
+	;;
+    "")
+	with_fp_cond=0
+	;;
+    *)
+	AC_MSG_ERROR([Unable to determine FLEN], [1])
+	;;
+    esac
+
+    case "$float_abi" in
+    soft)
+	abi_flen=0
+	;;
+    single)
+	AC_MSG_ERROR([glibc does not yet support the single floating-point ABI], [1])
+	;;
+    double)
+	abi_flen=64
+	;;
+    *)
+	AC_MSG_ERROR([Unable to determine floating-point ABI], [1])
+	;;
+    esac
+
+    case "$atomic" in
+    __riscv_atomic)
+	;;
+    *)
+	AC_MSG_ERROR([glibc requires the A extension], [1])
+	;;
+    esac
+
+    base_machine=riscv
+    machine=riscv/rv$xlen/$float_machine
+
+    AC_DEFINE_UNQUOTED([RISCV_ABI_XLEN], [$xlen])
+    AC_DEFINE_UNQUOTED([RISCV_ABI_FLEN], [$abi_flen])
+    ;;
+esac

Re: [PATCH v2 4/4] RISC-V: Use an autoconf template to produce `preconfigure'

[Palmer Dabbelt]

On Thu, 12 May 2022 15:29:11 PDT (-0700), macro@embecosm.com wrote:
> Avoid fiddling with autoconf internals and use AC_DEFINE_UNQUOTED to
> define macros in the configuration headers rather than handcoding an
> equivalent shell sequence with the use of the `as_echo' undocumented
> variable.
>
> Switch to using AC_MSG_ERROR rather than `echo' and `exit' directly for
> error handling.  Owing to the lack of any kind of error annotation it
> makes it difficult to spot the message in the flood in a parallel build
> and neither it is logged in `config.log'.
> ---
> Changes from v1:
>
> - Convert the script to an autoconf template and use AC_MSG_ERROR.
>
> - Also handle configuration macro definitions, using AC_DEFINE_UNQUOTED.
> ---
>  sysdeps/riscv/preconfigure    |   33 +++++++++++----------
>  sysdeps/riscv/preconfigure.ac |   64 ++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 82 insertions(+), 15 deletions(-)
>
> glibc-riscv-preconfigure-as-fn-error.diff
> Index: glibc/sysdeps/riscv/preconfigure
> ===================================================================
> --- glibc.orig/sysdeps/riscv/preconfigure
> +++ glibc/sysdeps/riscv/preconfigure
> @@ -1,3 +1,6 @@
> +# This file is generated from configure.ac by Autoconf.  DO NOT EDIT!
> +# Local preconfigure fragment for sysdeps/riscv
> +
>  case "$machine" in
>  riscv*)
>      xlen=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __riscv_xlen \(.*\)/\1/p'`
> @@ -9,8 +12,7 @@ riscv*)
>      64 | 32)
>  	;;
>      *)
> -	echo "Unable to determine XLEN" >&2
> -	exit 1
> +	as_fn_error 1 "Unable to determine XLEN" "$LINENO" 5
>  	;;
>      esac
>
> @@ -20,15 +22,13 @@ riscv*)
>  	with_fp_cond=1
>  	;;
>      32)
> -	echo "glibc does not yet support systems with the F but not D extensions" >&2
> -	exit 1
> +	as_fn_error 1 "glibc does not yet support systems with the F but not D extensions" "$LINENO" 5
>  	;;
>      "")
>  	with_fp_cond=0
>  	;;
>      *)
> -	echo "Unable to determine FLEN" >&2
> -	exit 1
> +	as_fn_error 1 "Unable to determine FLEN" "$LINENO" 5
>  	;;
>      esac
>
> @@ -37,31 +37,34 @@ riscv*)
>  	abi_flen=0
>  	;;
>      single)
> -	echo "glibc does not yet support the single floating-point ABI" >&2
> -	exit 1
> +	as_fn_error 1 "glibc does not yet support the single floating-point ABI" "$LINENO" 5
>  	;;
>      double)
>  	abi_flen=64
>  	;;
>      *)
> -	echo "Unable to determine floating-point ABI" >&2
> -	exit 1
> +	as_fn_error 1 "Unable to determine floating-point ABI" "$LINENO" 5
>  	;;
>      esac
>
>      case "$atomic" in
>      __riscv_atomic)
> -        ;;
> +	;;
>      *)
> -        echo "glibc requires the A extension" >&2
> -	exit 1
> +	as_fn_error 1 "glibc requires the A extension" "$LINENO" 5
>  	;;
>      esac
>
>      base_machine=riscv
>      machine=riscv/rv$xlen/$float_machine
>
> -    $as_echo "#define RISCV_ABI_XLEN $xlen" >>confdefs.h
> -    $as_echo "#define RISCV_ABI_FLEN $abi_flen" >>confdefs.h
> +    cat >>confdefs.h <<_ACEOF
> +#define RISCV_ABI_XLEN $xlen
> +_ACEOF
> +
> +    cat >>confdefs.h <<_ACEOF
> +#define RISCV_ABI_FLEN $abi_flen
> +_ACEOF
> +
>      ;;
>  esac
> Index: glibc/sysdeps/riscv/preconfigure.ac
> ===================================================================
> --- /dev/null
> +++ glibc/sysdeps/riscv/preconfigure.ac
> @@ -0,0 +1,64 @@
> +GLIBC_PROVIDES[]dnl See aclocal.m4 in the top level source directory.
> +# Local preconfigure fragment for sysdeps/riscv
> +
> +case "$machine" in
> +riscv*)
> +    xlen=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __riscv_xlen \(.*\)/\1/p'`
> +    flen=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __riscv_flen \(.*\)/\1/p'`
> +    float_abi=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | sed -n 's/^#define __riscv_float_abi_\([^ ]*\) .*/\1/p'`
> +    atomic=`$CC $CFLAGS $CPPFLAGS -E -dM -xc /dev/null | grep '#define __riscv_atomic' | cut -d' ' -f2`
> +
> +    case "$xlen" in
> +    64 | 32)
> +	;;
> +    *)
> +	AC_MSG_ERROR([Unable to determine XLEN], [1])
> +	;;
> +    esac
> +
> +    case "$flen" in
> +    64)
> +	float_machine=rvd
> +	with_fp_cond=1
> +	;;
> +    32)
> +	AC_MSG_ERROR([glibc does not yet support systems with the F but not D extensions], 1)
> +	;;
> +    "")
> +	with_fp_cond=0
> +	;;
> +    *)
> +	AC_MSG_ERROR([Unable to determine FLEN], [1])
> +	;;
> +    esac
> +
> +    case "$float_abi" in
> +    soft)
> +	abi_flen=0
> +	;;
> +    single)
> +	AC_MSG_ERROR([glibc does not yet support the single floating-point ABI], [1])
> +	;;
> +    double)
> +	abi_flen=64
> +	;;
> +    *)
> +	AC_MSG_ERROR([Unable to determine floating-point ABI], [1])
> +	;;
> +    esac
> +
> +    case "$atomic" in
> +    __riscv_atomic)
> +	;;
> +    *)
> +	AC_MSG_ERROR([glibc requires the A extension], [1])
> +	;;
> +    esac
> +
> +    base_machine=riscv
> +    machine=riscv/rv$xlen/$float_machine
> +
> +    AC_DEFINE_UNQUOTED([RISCV_ABI_XLEN], [$xlen])
> +    AC_DEFINE_UNQUOTED([RISCV_ABI_FLEN], [$abi_flen])
> +    ;;
> +esac

Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>

Thanks!

Re: [PATCH v2 0/4] Avoid relying on autoconf internals in `preconfigure' scripts

[Palmer Dabbelt]

On Thu, 12 May 2022 15:28:27 PDT (-0700), macro@embecosm.com wrote:
> Hi,
>
>  I have discovered that we have two kinds of `preconfigure' scripts,
> handwritten ones and autoconf-generated ones.  While it seems perfectly
> fine to have these scripts handwritten, some of them wire into autoconf
> internals or use an ad-hoc way to report errors.  The scripts for the
> C-SKY, m68k, MIPS and RISC-V targets fall into this category.
>
>  This small patch series addresses this problem by converting these
> scripts to autoconf templates with suitable adjustments, then processed
> with autoconf to re-create functionally equivalent almost identical
> scripts.
>
>  The RISC-V script has been verified to still correctly build the library.
> All have been visually inspected to ensure functional equivalence; changes
> are so minimal for this to be straightforward.  There's some noise though
> from a formatting fix to replace octets of spaces with tabs in the new
> templates.
>
>  OK to apply?

OK from me for the RISC-V bits (I'd had a draft up for the v1 waiting 
for me to build it, as that one hadn't said anything), but given that I 
probably broke them in the first place (and likely because I don't 
understand autoconf) that might not be the best ack ;)

Thanks!

>
>   Maciej