[2/2] RFC: build machoread.c in more cases

[2/2] RFC: build machoread.c in more cases

[Tom Tromey]

I noticed that machoread.c is only built for Darwin.

This seems wrong to me.  It seems like it should instead be built
whenever the needed BFD code is built.  This is what we do for
elfread.c.

This patch has two benefits.  First, it means building with a Darwin
target is more possible.  Second, it means that --enable-targets=all
will build machoread.c, leading to fewer chances for build regressions
there.  The is the case I ran into...

Comments?

Tom

	* config/i386/darwin.mh (NATDEPFILES): Remove machoread.o.
	* configure.ac: Check for Mach-O support in BFD.  Update
	CONFIG_OBS.
	* configure: Rebuild.
---
 gdb/config/i386/darwin.mh |    2 +-
 gdb/configure             |   52 +++++++++++++++++++++++++++++++++++++++++++++
 gdb/configure.ac          |    7 ++++++
 3 files changed, 60 insertions(+), 1 deletions(-)

diff --git a/gdb/config/i386/darwin.mh b/gdb/config/i386/darwin.mh
index cd7f248..190d56c 100644
--- a/gdb/config/i386/darwin.mh
+++ b/gdb/config/i386/darwin.mh
@@ -1,4 +1,4 @@
 # Host: IA86 running Darwin
 
-NATDEPFILES = fork-child.o machoread.o darwin-nat.o \
+NATDEPFILES = fork-child.o darwin-nat.o \
      i386-darwin-nat.o i386-nat.o amd64-nat.o darwin-nat-info.o
diff --git a/gdb/configure.ac b/gdb/configure.ac
index 26c88f4..0468f3b 100644
--- a/gdb/configure.ac
+++ b/gdb/configure.ac
@@ -2027,6 +2027,13 @@ if test $gdb_cv_var_elf = yes; then
   fi
 fi
 
+# Add macho support to GDB, but only if BFD includes it.
+GDB_AC_CHECK_BFD([for Mach-O support in BFD], gdb_cv_var_macho,
+                 [bfd_mach_o_lookup_command (NULL, 0, NULL)], mach-o.h)
+if test $gdb_cv_var_macho = yes; then
+  CONFIG_OBS="$CONFIG_OBS machoread.o"
+fi
+
 # Add any host-specific objects to GDB.
 CONFIG_OBS="${CONFIG_OBS} ${gdb_host_obs}"
 
-- 
1.7.7.6

Re: [2/2] RFC: build machoread.c in more cases

[Pedro Alves]

On 12/13/2012 07:53 PM, Tom Tromey wrote:
> I noticed that machoread.c is only built for Darwin.
> 
> This seems wrong to me.  It seems like it should instead be built
> whenever the needed BFD code is built.  This is what we do for
> elfread.c.
> 
> This patch has two benefits.  First, it means building with a Darwin
> target is more possible.  Second, it means that --enable-targets=all
> will build machoread.c, leading to fewer chances for build regressions
> there.  The is the case I ran into...
> 
> Comments?

Looks fine to me.

-- 
Pedro Alves

Re: [2/2] RFC: build machoread.c in more cases

[Tom Tromey]

>>>>> "Tom" == Tom Tromey <tromey@redhat.com> writes:

Tom> I noticed that machoread.c is only built for Darwin.
Tom> This seems wrong to me.  It seems like it should instead be built
Tom> whenever the needed BFD code is built.  This is what we do for
Tom> elfread.c.

I'm checking this in now.

Tom