[PATCH] regex: fix a SIZE_MAX macro redefinition warning

Since commit 56a1a3ab ("Silence GCC's \"cast of pointer to integer of a
different size\" warning", 26-10-2015), sparse has been issuing a macro
redefinition warning for the SIZE_MAX macro. However, gcc did not issue
any such warning.

After commit 56a1a3ab, in terms of the order of #includes and #defines,
the code looked something like:

  $ cat -n junk.c
       1	#include <stddef.h>
       2
       3	#define SIZE_MAX ((size_t) -1)
       4
       5	#include <stdint.h>
       6
       7	int main(int argc, char *argv[])
       8	{
       9		return 0;
      10	}
  $
  $ gcc junk.c
  $

However, if you compile that file with -Wsystem-headers, then it will
also issue a warning. Having set -Wsystem-headers in CFLAGS, using the
config.mak file, then (on cygwin):

  $ make compat/regex/regex.o
      CC compat/regex/regex.o
  In file included from /usr/lib/gcc/x86_64-pc-cygwin/4.9.3/include/stdint.h:9:0,
                   from compat/regex/regcomp.c:21,
                   from compat/regex/regex.c:77:
  /usr/include/stdint.h:362:0: warning: "SIZE_MAX" redefined
   #define SIZE_MAX (__SIZE_MAX__)
   ^
  In file included from compat/regex/regex.c:69:0:
  compat/regex/regex_internal.h:108:0: note: this is the location of the previous definition
   # define SIZE_MAX ((size_t) -1)
   ^
  $

The compilation of the compat/regex code is somewhat unusual in that the
regex.c file directly #includes the other c files (regcomp.c, regexec.c
and regex_internal.c). Commit 56a1a3ab added an #include of <stdint.h>
to the regcomp.c file, which results in the redefinition, since this is
included after the regex_internal.h header. This header file contains a
'fallback' definition for SIZE_MAX, in order to support systems which do
not have the <stdint.h> header (the HAVE_STDINT_H macro is not defined).

In order to suppress the warning, we remove the #include of <stdint.h>
from regcomp.c and set the HAVE_STDINT_H macro, using the regex.o build
rule within the Makefile, to ensure that <stdint.h> is #included from
the regex_internal.h header.

Signed-off-by: Ramsay Jones <redacted>
---

Hi Junio,

About seven months ago, sparse started complaining but, because gcc wasn't, I
just assumed this was a sparse bug. I put it on my sparse TODO list and pretty
much forgot about it. Tonight I decided to take a quick look and was a bit
surprised by what I found. ;-)

I spent some time worrying about the 'unconditional inclusion' of <stdint.h>
(there used to be systems that didn't supply that header that we wanted to
support), but I suspect that is now a non-issue. In any event, I don't
think this patch makes things any worse.

[Just a flavour of the rabbit holes I went down; HP-NONSTOP seems to supply
both <inttypes.h> and <stdint.h> but it doesn't supply the intptr_t and
uintptr_t types (ie it defines NO_INTPTR_T). So, I checked the C99 and C11
standards and, blow me down, but the standard states that those types are
_optional_. (C99 7.18.1.4) :-D ]

ATB,
Ramsay Jones

 Makefile               | 2 +-
 compat/regex/regcomp.c | 2 --
 2 files changed, 1 insertion(+), 3 deletions(-)

diff --git a/Makefile b/Makefile
index 0d59718..3f6c70a 100644
--- a/Makefile
+++ b/Makefile

@@ -1987,7 +1987,7 @@ endif
 
 ifdef NO_REGEX
 compat/regex/regex.sp compat/regex/regex.o: EXTRA_CPPFLAGS = \
-	-DGAWK -DNO_MBSUPPORT
+	-DGAWK -DNO_MBSUPPORT -DHAVE_STDINT_H
 endif
 
 ifdef USE_NED_ALLOCATOR

diff --git a/compat/regex/regcomp.c b/compat/regex/regcomp.c
index fba5986..d8bde06 100644
--- a/compat/regex/regcomp.c
+++ b/compat/regex/regcomp.c

@@ -18,8 +18,6 @@
    Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
    02110-1301 USA.  */
 
-#include <stdint.h>
-
 static reg_errcode_t re_compile_internal (regex_t *preg, const char * pattern,
 					  size_t length, reg_syntax_t syntax);
 static void re_compile_fastmap_iter (regex_t *bufp,

-- 
2.8.0

Hi Ramsay,

thanks for working on this!

On Sat, 4 Jun 2016, Ramsay Jones wrote:

Maybe a comment here, something like "the fallback regex implementation
*requires* stdint.h"?

Other than that, I think this patch is an improvement.

Thanks again,
Dscho

On 05/06/16 08:15, Johannes Schindelin wrote:

Hi Ramsay,

thanks for working on this!

On Sat, 4 Jun 2016, Ramsay Jones wrote:

quoted

diff --git a/Makefile b/Makefile
index 0d59718..3f6c70a 100644
--- a/Makefile
+++ b/Makefile

@@ -1987,7 +1987,7 @@ endif
 
 ifdef NO_REGEX
 compat/regex/regex.sp compat/regex/regex.o: EXTRA_CPPFLAGS = \
-	-DGAWK -DNO_MBSUPPORT
+	-DGAWK -DNO_MBSUPPORT -DHAVE_STDINT_H

Maybe a comment here, something like "the fallback regex implementation
*requires* stdint.h"?

The original version of this patch looked like this:

  diff --git a/compat/regex/regcomp.c b/compat/regex/regcomp.c
  index fba5986..d8bde06 100644
  --- a/compat/regex/regcomp.c
  +++ b/compat/regex/regcomp.c
  @@ -18,8 +18,6 @@
      Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
      02110-1301 USA.  */
   
  -#include <stdint.h>
  -
   static reg_errcode_t re_compile_internal (regex_t *preg, const char * pattern,
   					  size_t length, reg_syntax_t syntax);
   static void re_compile_fastmap_iter (regex_t *bufp,
  diff --git a/compat/regex/regex.c b/compat/regex/regex.c
  index 6aaae00..5cb23e5 100644
  --- a/compat/regex/regex.c
  +++ b/compat/regex/regex.c
  @@ -60,6 +60,7 @@
      GNU regex allows.  Include it before <regex.h>, which correctly
      #undefs RE_DUP_MAX and sets it to the right value.  */
   #include <limits.h>
  +#include <stdint.h>
   
   #ifdef GAWK
   #undef alloca

So, just move the unconditional inclusion to the start of the compilation
unit root file, before the #include of the regex_internal.h header.

In some ways this is a better fix, because it makes it clear that, currently,
the compat/regex code requires <stdint.h>. This would remove the need for
such a comment.

This effectively makes the conditional inclusion of <stdint.h>, and the SIZE_MAX
fallback, in regex_internal.h dead code. (The C99 standard _requires_ the
definition of SIZE_MAX in <stdint.h>, thankfully! ;-). So, I was tempted to
remove them as part of the patch. However, I also wanted to minimize the changes
to the regex code, just in case we ever wanted to re-import a newer version
from upstream. Setting HAVE_STDINT_H seemed like a good solution, but maybe the
first patch would be more honest?

As I said earlier, I was a little concerned about the 'unconditional' aspect of
the inclusion of <stdint.h>. At one time we wanted to support systems that didn't
have <stdint.h> (or didn't have <inttypes.h> but did have <stdint.h>). However,
it has been several months and we have not heard anyone scream, so ...

It is slightly amusing that the reason you #included <stdint.h> was to get the
definition of 'intptr_t' and the C standard states that this type is optional.
In practice, I suspect that the number of platforms which do not define 'intptr_t'
and 'uintptr_t' in the <stdint.h> header is rather small.

Having said that ... If I'm reading the code/config correctly, HP-NONSTOP would
be failing to compile at the moment. (Although it has <stdint.h>, it does not
define 'intptr_t' - ie it defines 'NO_REGEX = YesPlease' and 'NO_INTPTR_T = \
UnfortunatelyYes').

Other than that, I think this patch is an improvement.

Thanks. What do you think of replacing it with the original patch (above)?

ATB,
Ramsay Jones

Hi Ramsay,

On Sun, 5 Jun 2016, Ramsay Jones wrote:

quoted

Other than that, I think this patch is an improvement.

Thanks. What do you think of replacing it with the original patch
(above)?

That would work for me, too.

Thanks,
Dscho