Re: HPUX 11 "size_t" fixinc problems

Re: HPUX 11 "size_t" fixinc problems

[Steve Ellcey]

I wonder if it would be a good idea to also add "STMP_FIXPROTO=" to t-pa
and t-pa64 on the mainline.  At least on HP-UX 11.* I have found no
need to run fixproto as all the headers have the proper prototypes but
fixproto still does some unneccessary fixes when it finds functions with
wrong/obsolete prototypes that are under ifdefs that ensure they will
never be seen by GCC anyway.

This is working fine for me on HP-UX 11.* systems and reduces the number
of headers that get modified, I have not tried it on 10.*.

Steve Ellcey
sje@cup.hp.com

> Therefore, it is my intention to remove all the empty replacement
> non-fixes and wait to see what problems arise.  (I don't have 
> enough of a variety of systems to do it any other way.)  So, I
> have also posted this to the GCC list.
> 
> SO EVERYONE BE ADVISED:  I am removing the protection many headers
> used to have against mis-fixes from fixincludes.  I will put them
> back only if the fixincl program cannot be fixed itself.
> 
> Um, just to be clear:  mainline only, not 3.0 branch.
> 
> Cheers,
> 	Bruce

Re: HPUX 11 "size_t" fixinc problems

[John David Anglin]

> Excellent.  I applied this to the 3.0 branch before Friday's freeze.
> I intend to apply a slightly different one to the main line,
> barring complaints.  On the theory that all of the "empty replacement"
> fixes are really fixes for the broken fixincludes script, my
> _guess_ is that they are not needed any more because my version
> doesn't make mistakes.  ;-)

I have tested the following under hpux 11.00 and 11.11.  No fixes were
applied to any of the sys/ki*.h headers in either case.  So, I am happy
with the removal of the the AAA_ki* fixes on the main.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)


2001-12-17  John David Anglin  <dave@hiauly1.hia.nrc.ca>

        * inclhack.def (AAA_ki, AAA_ki_calls, AAA_ki_defs, AAA_ki_iface,
	AAA_time): Delete.
        (hpux_size_t): Revise select and fix for HP-UX 11.11.
        (hpux11_vsnprintf): Likewise.

--- inclhack.def.orig	Thu Dec  6 21:55:49 2001
+++ inclhack.def	Sun Dec 16 12:51:08 2001
@@ -39,50 +39,6 @@
 
 
 /*
- *  Purge some HP-UX 11 files that are only broken after they are "fixed".
- */
-fix = {
-    hackname = AAA_ki;
-    files    = sys/ki.h;
-    select   = '11.00 HP-UX LP64';
-    replace; /* empty replacement -> no fixing the file */
-};
-
-
-/*
- *  Purge some HP-UX 11 files that are only broken after they are "fixed".
- */
-fix = {
-    hackname = AAA_ki_calls;
-    files    = sys/ki_calls.h;
-    select   = 'kthread_create_caller_t';
-    replace; /* empty replacement -> no fixing the file */
-};
-
-
-/*
- *  Purge some HP-UX 11 files that are only broken after they are "fixed".
- */
-fix = {
-    hackname = AAA_ki_defs;
-    files    = sys/ki_defs.h;
-    select   = 'Kernel Instrumentation Definitions';
-    replace; /* empty replacement -> no fixing the file */
-};
-
-
-/*
- *  Purge some HP-UX 11 files that are only broken after they are "fixed".
- */
-fix = {
-    hackname = AAA_ki_iface;
-    files    = sys/ki_iface.h;
-    select   = 'These definitions are for HP Internal developers';
-    replace; /* empty replacement -> no fixing the file */
-};
-
-
-/*
  * Tru64 UNIX V4.0F/V5.1 <standards.h> defines _NO_PROTO and _NONSTD_TYPES
  * correctly for GCC, but strict_ansi_not breaks it.
  */
@@ -94,16 +50,6 @@
 };
 
 
-/*
- *  Purge some HP-UX 11 files that are only broken after they are "fixed".
- */
-fix = {
-    hackname = AAA_time;
-    files    = sys/time.h;
-    select   = '11.0 and later representation of ki time';
-    replace; /* empty replacement -> no fixing the file */
-};
-
 /* And now, the real fixes, replacement text fixes first: */
 
 /*
@@ -1222,12 +1168,11 @@
  */
 fix = {
     hackname  = hpux11_size_t;
-    mach      = "*-hp-hpux*";
-    select    = "^#define __size_t size_t";
+    mach      = "*-hp-hpux11*";
+    select    = "__size_t";
 
     c_fix     = format;
     c_fix_arg = "_hpux_size_t";
-    c_fix_arg = "__size_t";
 
     test_text =
     "#define __size_t size_t\n"
@@ -1264,17 +1209,17 @@
 
 
 /*
- *  Fix hpux 11.00 broken vsnprintf declaration
+ *  Fix hpux 11.00 broken vsnprintf declaration.
  */
 fix = {
     hackname = hpux11_vsnprintf;
     files    = stdio.h;
-    select   = 'extern int vsnprintf\(char \*, __size_t, const char \*,'
+    select   = 'extern int vsnprintf\(char \*, .*_size_t, const char \*,'
                                      ' __va__list\);';
 
     c_fix     = format;
-    c_fix_arg = "extern int vsnprintf(char *, __size_t, const char *,"
-                                     " __va_list);";
+    c_fix_arg = __va_list;
+    c_fix_arg = __va__list;
 
     test_text = 'extern int vsnprintf(char *, __size_t, const char *,'
                                      ' __va__list);';

Re: HPUX 11 "size_t" fixinc problems

[Bruce Korb]

John David Anglin wrote:

> I have run this through complete bootstraps and checks on hppa2.0w-hp-hpux11.00
> and hppa2.0w-hp-hpux11.11 with no regressions.
> 
> Dave

> --- inclhack.def.orig   Thu Sep 13 12:05:12 2001
> +++ inclhack.def        Fri Nov 30 15:53:00 2001
> @@ -94,16 +94,6 @@
>  };
> 
> 
> -/*
> - *  Purge some HP-UX 11 files that are only broken after they are "fixed".
> - */
> -fix = {
> -    hackname = AAA_time;
> -    files    = sys/time.h;
> -    select   = '11.0 and later representation of ki time';
> -    replace; /* empty replacement -> no fixing the file */
> -};
> -
>  /* And now, the real fixes, replacement text fixes first: */

Hi Dave,

Excellent.  I applied this to the 3.0 branch before Friday's freeze.
I intend to apply a slightly different one to the main line,
barring complaints.  On the theory that all of the "empty replacement"
fixes are really fixes for the broken fixincludes script, my
_guess_ is that they are not needed any more because my version
doesn't make mistakes.  ;-)

Therefore, it is my intention to remove all the empty replacement
non-fixes and wait to see what problems arise.  (I don't have 
enough of a variety of systems to do it any other way.)  So, I
have also posted this to the GCC list.

SO EVERYONE BE ADVISED:  I am removing the protection many headers
used to have against mis-fixes from fixincludes.  I will put them
back only if the fixincl program cannot be fixed itself.

Um, just to be clear:  mainline only, not 3.0 branch.

Cheers,
	Bruce