Rework CC/CC_STAGE0 handling in `configure.ac`

Rather than using the non-standard/idiomatic `--with-{gcc,clang}=...`
scheme use the `CC=...` style scheme.

The basic idea is to have Autoconf's CC/CFLAG/CPPFLAG apply to
stage{1,2,3}, while having a separate _STAGE0 set of env-vars
denote the bootstrap-toolchain flags/programs.

This should be simpler, less confusing, and somewhat more in line with
Autoconf's idioms (allowing us to reuse more of Autoconf rather than
(re)inventing our own confusing non-standard m4 macros to do stuff that
Autoconf could almost do already for us)

Morever, expose CC_STAGE0 as a so-called "precious" variable.

So now we can better control which bootstrapping gcc is used
(by default the one used by the stage0 ghc, unless CC_STAGE0 is
overriden)

```
Some influential environment variables:
  CC_STAGE0   C compiler command (bootstrap)
  CC          C compiler command
  CFLAGS      C compiler flags
  ...

Use these variables to override the choices made by `configure' or to
help it to find libraries and programs with nonstandard names/locations.
```

Test Plan: I've tested that cross-compiling with
`--target=powerpc-linux-gnu` still works, and tried a few variants of
settting `CC=` and `CC_STAGE0=`; `./validate` passed as well

Reviewers: erikd, austin, bgamari, simonmar

Reviewed By: simonmar

Subscribers: Phyx, thomie

Differential Revision: https://phabricator.haskell.org/D2078

8 files changed, 42 insertions, 55 deletions

diff --git a/aclocal.m4 b/aclocal.m4
index 6403748c7f..477da3f05d 100644
--- a/aclocal.m4
+++ b/aclocal.m4
@@ -471,7 +471,7 @@ AC_DEFUN([FP_SETTINGS],
         SettingsWindresCommand="\$topdir/../${mingw_bin_prefix}windres.exe"
         SettingsTouchCommand='$topdir/bin/touchy.exe'
     else
-        SettingsCCompilerCommand="$WhatGccIsCalled"
+        SettingsCCompilerCommand="$CC"
         SettingsHaskellCPPCommand="$HaskellCPPCmd"
         SettingsHaskellCPPFlags="$HaskellCPPArgs"
         SettingsLdCommand="$LdCmd"
@@ -2067,38 +2067,6 @@ AC_DEFUN([FIND_GHC_BOOTSTRAP_PROG],[
 ])
 
 
-# FIND_GCC()
-# --------------------------------
-# Finds where gcc is
-#
-# $1 = the variable to set
-# $2 = the with option name
-# $3 = the command to look for
-AC_DEFUN([FIND_GCC],[
-    if test "$windows" = YES
-    then
-        $1="$CC"
-    else
-        FP_ARG_WITH_PATH_GNU_PROG_OPTIONAL([$1], [$2], [$3])
-        # fallback to CC if set and no --with-$2=... was used
-        if test -z "$With_$2" -a -n "$CC"
-        then
-            With_$2="$CC"
-            $1="$CC"
-        # From Xcode 5 on/, OS X command line tools do not include gcc
-        # anymore. Use clang.
-        elif test -z "$$1"
-        then
-            FP_ARG_WITH_PATH_GNU_PROG_OPTIONAL([$1], [clang], [clang])
-        fi
-        if test -z "$$1"
-        then
-            AC_MSG_ERROR([cannot find $3 nor clang in your PATH])
-        fi
-    fi
-    AC_SUBST($1)
-])
-
 AC_DEFUN([MAYBE_OVERRIDE_STAGE0],[
   if test ! -z "$With_$1" -a "$CrossCompiling" != "YES"; then
       AC_MSG_NOTICE([Not cross-compiling, so --with-$1 also sets $2])
@@ -2132,13 +2100,13 @@ AC_ARG_WITH(hs-cpp,
 
     # We can't use $CPP here, since HS_CPP_CMD is expected to be a single
     # command (no flags), and AC_PROG_CPP defines CPP as "/usr/bin/gcc -E".
-    HS_CPP_CMD=$WhatGccIsCalled
+    HS_CPP_CMD=$CC
 
     SOLARIS_GCC_CPP_BROKEN=NO
     SOLARIS_FOUND_GOOD_CPP=NO
     case $host in
         i386-*-solaris2)
-        GCC_MAJOR_MINOR=`$WhatGccIsCalled --version|grep "gcc (GCC)"|cut -d ' ' -f 3-3|cut -d '.' -f 1-2`
+        GCC_MAJOR_MINOR=`$CC --version|grep "gcc (GCC)"|cut -d ' ' -f 3-3|cut -d '.' -f 1-2`
         if test "$GCC_MAJOR_MINOR" != "3.4"; then
           # this is not 3.4.x release so with broken CPP
           SOLARIS_GCC_CPP_BROKEN=YES
diff --git a/bindisttest/ghc.mk b/bindisttest/ghc.mk
index 4b211064b1..cf2e481efd 100644
--- a/bindisttest/ghc.mk
+++ b/bindisttest/ghc.mk
@@ -38,7 +38,7 @@ test_bindist:
 ifeq "$(Windows_Host)" "YES"
 	mv bindisttest/a/b/c/$(BIN_DIST_NAME) $(BIN_DIST_INST_DIR)
 else
-	cd bindisttest/a/b/c/$(BIN_DIST_NAME) && ./configure --prefix=$(TOP)/$(BIN_DIST_INST_DIR) --with-gcc="$(WhatGccIsCalled)"
+	cd bindisttest/a/b/c/$(BIN_DIST_NAME) && ./configure --prefix=$(TOP)/$(BIN_DIST_INST_DIR) CC="$(CC)"
 	cd bindisttest/a/b/c/$(BIN_DIST_NAME) && $(MAKE) install
 endif
 ifeq "$(GhcProfiled)" "NO"
diff --git a/configure.ac b/configure.ac
index dd3236c24e..c5260ce609 100644
--- a/configure.ac
+++ b/configure.ac
@@ -33,6 +33,11 @@ fi
 
 AC_SUBST([CONFIGURE_ARGS], [$ac_configure_args])
 
+dnl this makes sure `./configure --target=<cross-compile-target>`
+dnl works as expected, since we're slightly modifying how Autoconf
+dnl interprets build/host/target and how this interacts with $CC tests
+test -n "$target_alias" && ac_tool_prefix=$target_alias-
+
 dnl ----------------------------------------------------------
 dnl ** Find unixy sort and find commands,
 dnl ** which are needed by FP_SETUP_PROJECT_VERSION
@@ -107,6 +112,11 @@ AC_ARG_ENABLE(tarballs-autodownload,
         TarballsAutodownload=YES,
         TarballsAutodownload=NO
 )
+
+dnl CC_STAGE0 is like the "previous" variable CC (inherited by CC_STAGE[123])
+dnl but instead used by stage0 for bootstrapping stage1
+AC_ARG_VAR(CC_STAGE0, [C compiler command (bootstrap)])
+
 if test "$WithGhc" != ""; then
   FPTOOLS_GHC_VERSION([GhcVersion], [GhcMajVersion], [GhcMinVersion], [GhcPatchLevel])dnl
 
@@ -121,7 +131,11 @@ if test "$WithGhc" != ""; then
   GhcMinVersion2=`echo "$GhcMinVersion" | sed 's/^\\(.\\)$/0\\1/'`
   GhcCanonVersion="$GhcMajVersion$GhcMinVersion2"
 
-  BOOTSTRAPPING_GHC_INFO_FIELD([CC_STAGE0],[C compiler command])
+  dnl infer CC_STAGE0 from `ghc --info` unless explicitly set by user
+  if test -z "$CC_STAGE0"; then
+    BOOTSTRAPPING_GHC_INFO_FIELD([CC_STAGE0],[C compiler command])
+  fi
+
   BOOTSTRAPPING_GHC_INFO_FIELD([LD_STAGE0],[ld command])
   BOOTSTRAPPING_GHC_INFO_FIELD([AR_STAGE0],[ar command])
   BOOTSTRAPPING_GHC_INFO_FIELD([AR_OPTS_STAGE0],[ar flags])
@@ -432,13 +446,19 @@ AC_SUBST(TargetPlatformFull)
 
 dnl ** Which gcc to use?
 dnl --------------------------------------------------------------
-FIND_GCC([WhatGccIsCalled], [gcc], [gcc])
-CC="$WhatGccIsCalled"
-export CC
 
-# If --with-gcc was used, and we're not cross-compiling, then it also
-# applies to the stage0 compiler.
-MAYBE_OVERRIDE_STAGE0([gcc],[CC_STAGE0])
+AC_ARG_WITH([gcc],
+        AS_HELP_STRING([--with-gcc=ARG], [Use ARG as the path to gcc (obsolete, use CC=ARG instead) [default=autodetect]]),
+        AC_MSG_ERROR([--with-gcc=$withval is obsolete (use './configure CC=$withval' or 'CC=$withval ./configure' instead)]))
+
+AC_ARG_WITH([clang],
+        AS_HELP_STRING([--with-clang=ARG], [Use ARG as the path to gcc (obsolete, use CC=ARG instead) [default=autodetect]]),
+        AC_MSG_ERROR([--with-clang=$withval is obsolete (use './configure CC=$withval' or 'CC=$withval ./configure' instead)]))
+
+dnl detect compiler (prefer gcc over clang) and set $CC (unless CC already set),
+dnl later CC is copied to CC_STAGE{1,2,3}
+AC_PROG_CC([gcc clang])
+
 MAYBE_OVERRIDE_STAGE0([ar],[AR_STAGE0])
 
 dnl make extensions visible to allow feature-tests to detect them lateron
@@ -1155,7 +1175,8 @@ else
 fi
 
 echo ["\
-   Using $CompilerName       : $WhatGccIsCalled
+   Using (for bootstrapping) : $CC_STAGE0
+   Using $CompilerName       : $CC
       which is version       : $GccVersion
    Building a cross compiler : $CrossCompiling
    hs-cpp       : $HaskellCPPCmd
diff --git a/distrib/configure.ac.in b/distrib/configure.ac.in
index c0610c632e..fdd9fd92d6 100644
--- a/distrib/configure.ac.in
+++ b/distrib/configure.ac.in
@@ -59,9 +59,7 @@ XCODE_VERSION()
 
 dnl ** Which gcc to use?
 dnl --------------------------------------------------------------
-FIND_GCC([WhatGccIsCalled], [gcc], [gcc])
-CC="$WhatGccIsCalled"
-export CC
+AC_PROG_CC([gcc clang])
 
 # --with-hs-cpp/--with-hs-cpp-flags
 FP_CPP_CMD_WITH_ARGS(HaskellCPPCmd, HaskellCPPArgs)
diff --git a/includes/ghc.mk b/includes/ghc.mk
index 31b26508f6..d08ecfbc07 100644
--- a/includes/ghc.mk
+++ b/includes/ghc.mk
@@ -177,7 +177,7 @@ includes_GHCCONSTANTS_HASKELL_EXPORTS = includes/dist-derivedconstants/header/GH
 
 INSTALL_LIBS += $(includes_GHCCONSTANTS_HASKELL_VALUE)
 
-DERIVE_CONSTANTS_FLAGS += --gcc-program "$(WhatGccIsCalled)"
+DERIVE_CONSTANTS_FLAGS += --gcc-program "$(CC)"
 DERIVE_CONSTANTS_FLAGS += $(addprefix --gcc-flag$(space),$(includes_CC_OPTS) -fcommon)
 DERIVE_CONSTANTS_FLAGS += --nm-program "$(NM)"
 ifneq "$(OBJDUMP)" ""
diff --git a/mk/config.mk.in b/mk/config.mk.in
index 5d5e581a43..b5d910b841 100644
--- a/mk/config.mk.in
+++ b/mk/config.mk.in
@@ -509,11 +509,10 @@ SUPPORTS_THIS_UNIT_ID = @SUPPORTS_THIS_UNIT_ID@
 #-----------------------------------------------------------------------------
 # C compiler
 #
-# NB. Don't override $(WhatGccIsCalled) using build.mk,  re-configure using
-# the flag --with-gcc=<blah> instead.  The reason is that the configure script
+# NB. Don't override $(CC) using build.mk,  re-configure using
+# the flag CC=<blah> instead.  The reason is that the configure script
 # needs to know which gcc you're using in order to perform its tests.
 
-WhatGccIsCalled       = @WhatGccIsCalled@
 GccVersion            = @GccVersion@
 ifeq "$(phase)" "0"
 CrossCompilePrefix    =
@@ -527,18 +526,19 @@ GccLT34         = @GccLT34@
 GccLT46         = @GccLT46@
 GccIsClang      = @GccIsClang@
 
-CC              = $(WhatGccIsCalled)
+CC              = @CC@
 CC_STAGE0       = @CC_STAGE0@
 CC_STAGE1       = $(CC)
 CC_STAGE2       = $(CC)
 CC_STAGE3       = $(CC)
 
-AS              = $(WhatGccIsCalled)
+AS              = @CC@
 AS_STAGE0       = @CC_STAGE0@
 AS_STAGE1       = $(AS)
 AS_STAGE2       = $(AS)
 AS_STAGE3       = $(AS)
 
+# why no LD=@LD@ ?
 LD_STAGE0       = @LD_STAGE0@
 LD_STAGE1       = $(LD)
 LD_STAGE2       = $(LD)
diff --git a/rts/ghc.mk b/rts/ghc.mk
index f57f3dacd4..d7a46f909a 100644
--- a/rts/ghc.mk
+++ b/rts/ghc.mk
@@ -538,7 +538,7 @@ ifeq "$(TargetOS_CPP)" "darwin"
 # Darwin has a flag to tell dtrace which cpp to use.
 # Unfortunately, this isn't supported on Solaris (See Solaris Dynamic Tracing
 # Guide, Chapter 16, for the configuration variables available on Solaris)
-DTRACE_FLAGS = -x cpppath=$(WhatGccIsCalled)
+DTRACE_FLAGS = -x cpppath=$(CC)
 endif
 
 DTRACEPROBES_SRC = rts/RtsProbes.d
diff --git a/rules/shell-wrapper.mk b/rules/shell-wrapper.mk
index f674c0dfc2..eb7d8f13ab 100644
--- a/rules/shell-wrapper.mk
+++ b/rules/shell-wrapper.mk
@@ -46,7 +46,7 @@ $$($1_$2_INPLACE_WRAPPER): $$($1_$2_INPLACE)
 	echo 'datadir="$$(TOP)/$$(INPLACE_LIB)"'                             >> $$@
 	echo 'bindir="$$(TOP)/$$(INPLACE_BIN)"'                              >> $$@
 	echo 'topdir="$$(TOP)/$$(INPLACE_TOPDIR)"'                           >> $$@
-	echo 'pgmgcc="$$(WhatGccIsCalled)"'                                  >> $$@
+	echo 'pgmgcc="$$(CC)"'                                               >> $$@
 	$$($1_$2_SHELL_WRAPPER_EXTRA)
 	$$($1_$2_INPLACE_SHELL_WRAPPER_EXTRA)
 ifeq "$$(DYNAMIC_GHC_PROGRAMS)" "YES"