[rfc] configury: CC_HAS_UINTPTR_T

[rfc] configury: CC_HAS_UINTPTR_T

[Michael Elizabeth Chastain]

This is my first 'configure' patch, so I don't expect it to pass on
the first draft.  If there's anything I need to know about this
kind of stuff, tell me please.

This patch just adds CC_HAS_UINTPTR_T.  After this patch, I'll write
patches to procfs.c and gdb_thread_db.h to use the new variable, so that
they can put out a more informative compile error than the dreck that
they cough up right now.  I've seen several reports from the field where
this actually happens (pr gdb/660 is one case).

If you like CC_HAS_UINTPTR_T, but you don't like the idea of:

  #if !CC_HAS_UINTPTR_T
  #error 'Your glibc is so 20th century.  Bye bye!'
  #endif

... then I hope this patch is okay with you.  The next patch is the
place to debate what actually happens when CC_HAS_UINTPTR_T is not
defined.

I don't show the diffs to 'configure' because there's 3000 lines of
them.  I regenerated my 'configure' with a stock autoconf 2.13
from ftp.gnu.org.

Testing: I tested this by running it on native i686-pc-linux-gnu.
I played with /usr/include/stdint.h so that I could observe both
code paths.

Comments?

Michael C

2003-02-25  Michael Chastain  <mec@shout.net>

	* configure.in: New variable CC_HAS_UINTPTR_T.
	* config.in: Likewise.
	* configure: Regenerated.

Index: config.in
===================================================================
RCS file: /cvs/src/src/gdb/config.in,v
retrieving revision 1.53
diff -u -r1.53 config.in
--- config.in	14 Feb 2003 20:07:00 -0000	1.53
+++ config.in	25 Feb 2003 22:08:37 -0000
@@ -111,6 +111,9 @@
 /* Define if ioctl argument PIOCSET is available. */
 #undef HAVE_PROCFS_PIOCSET
 
+/* Define if the C library supports "uintptr_t" type. */
+#undef CC_HAS_UINTPTR_T
+
 /* Define if the `long long' type works.  */
 #undef CC_HAS_LONG_LONG
 
Index: configure.in
===================================================================
RCS file: /cvs/src/src/gdb/configure.in,v
retrieving revision 1.125
diff -u -r1.125 configure.in
--- configure.in	20 Feb 2003 23:38:58 -0000	1.125
+++ configure.in	25 Feb 2003 22:08:46 -0000
@@ -689,6 +689,23 @@
   fi
 fi
 
+dnl See if C library supports "uintptr_t" type.
+dnl AC_CHECK_TYPE is not enough because this type lives in a header file.
+
+AC_MSG_CHECKING(for uintptr_t in C library)
+AC_CACHE_VAL(gdb_cv_c_uintptr_t,
+[AC_TRY_RUN([
+#include <stdint.h>
+int main () {
+  uintptr_t foo = 0;
+  return foo;
+}],
+gdb_cv_c_uintptr_t=yes, gdb_cv_c_uintptr_t=no, gdb_cv_c_uintptr_t=no)])
+AC_MSG_RESULT($gdb_cv_c_uintptr_t)
+if test $gdb_cv_c_uintptr_t = yes; then
+  AC_DEFINE(CC_HAS_UINTPTR_T)
+fi
+
 dnl See if compiler supports "long long" type.
 
 AC_MSG_CHECKING(for long long support in compiler)

Re: [rfc] configury: CC_HAS_UINTPTR_T

[Daniel Jacobowitz]

On Tue, Feb 25, 2003 at 05:21:41PM -0600, Michael Elizabeth Chastain wrote:
> This is my first 'configure' patch, so I don't expect it to pass on
> the first draft.  If there's anything I need to know about this
> kind of stuff, tell me please.
> 
> This patch just adds CC_HAS_UINTPTR_T.  After this patch, I'll write
> patches to procfs.c and gdb_thread_db.h to use the new variable, so that
> they can put out a more informative compile error than the dreck that
> they cough up right now.  I've seen several reports from the field where
> this actually happens (pr gdb/660 is one case).

Could you call it something different?  Autoconfism for this would be
HAVE_UINTPTR_T.  You say yourself that it's a property of the C
library, not of the CC.

> +dnl See if C library supports "uintptr_t" type.
> +dnl AC_CHECK_TYPE is not enough because this type lives in a header file.
> +
> +AC_MSG_CHECKING(for uintptr_t in C library)
> +AC_CACHE_VAL(gdb_cv_c_uintptr_t,
> +[AC_TRY_RUN([

No no.  Use AC_TRY_COMPILE; never run a test that you don't actually
need runtime output for.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [rfc] configury: CC_HAS_UINTPTR_T

[Michael Elizabeth Chastain]

Hi Daniel,

mec> Could you call it something different?  Autoconfism for this would be
mec> HAVE_UINTPTR_T.  You say yourself that it's a property of the C
mec> library, not of the CC.

No problem.

mec>  +[AC_TRY_RUN([
drow> No no.  Use AC_TRY_COMPILE; never run a test that you don't actually
drow> need runtime output for.

I suppose that's a cross issue (because we can cross-compile but not
cross-run).

I originally wrote this with AC_TRY_COMPILE but it failed, because
AC_TRY_COMPILE wraps an "int main () { ... }" around its argument.

Err wait, two minutes with 'info automake' and I am enlightened.
Doh!

Thanks,

Michael C