Re: duplicate fixinc files

Re: duplicate fixinc files

[Marc Girod]

Hello Bruce,

Thanks for your reply.

>>>>> "BK" == Bruce Korb <korb@datadesign.com> writes:

BK> You are more than welcome to play with it on your box to give us a
BK> little more of a confidence base (if you are using a non-GNU-libc
BK> system):

BK> 	cvs get -rfixincl-branch egcs/gcc

OK. I built cvs on my hppa2.0 hp-ux 10.20 system, and did this.
Hopefully not too late for you.
I then reconfigured and built the gcc part (using autoconf 2.13 for
the first time and not touching the makefiles as recommended in the
FAQ for the older autoconf, or so I thought). Got a lot of errors.
Make clean, touch, configure and make.

BK> then verify that the ${builddir}/gcc/include is essentially
BK> the same as the non-fixincl-branch version.

And superficially it looks the same (no time to be more precise now).
I'll check deeper later. Tell me if there is something specific you
want me to look at.

BK> The reason for the non-GNU-libc system is that the
BK> fixincludes stuff is a noop on GNU-libc systems.

I'll not do it on my Linux Red Hat 5.2 box.

Best Regards!
Marc

-- 
Marc Girod                Hiomo 5/1          Voice:  +358-9-511 23746
Nokia Telecommunications  P.O. Box 320       Mobile: +358-40-569 7954
NWS/NMS/NMS for Data      00045 NOKIA Group  Fax:    +358-9-511 23580
                          Finland            marc.girod@ntc.nokia.com

Re: duplicate fixinc files

[Jeffrey A Law]

  In message < 1ybtk7tkh0.fsf@sinkku.ntc.nokia.com >you write:
  > OK. I built cvs on my hppa2.0 hp-ux 10.20 system, and did this.
  > Hopefully not too late for you.
  > I then reconfigured and built the gcc part (using autoconf 2.13 for
  > the first time and not touching the makefiles as recommended in the
  > FAQ for the older autoconf, or so I thought). Got a lot of errors.
  > Make clean, touch, configure and make.
  > 
  > BK> then verify that the ${builddir}/gcc/include is essentially
  > BK> the same as the non-fixincl-branch version.
  > 
  > And superficially it looks the same (no time to be more precise now).
  > I'll check deeper later. Tell me if there is something specific you
  > want me to look at.
Excellent!

jeff

Re: duplicate fixinc files

[Marc Girod]

>>>>> "JL" == Jeffrey A Law <law@hurl.cygnus.com> writes:

>> Do you want the diffs?
JL> Yes please.


egcs> find gcc/include | grep -vE '(gcc/include$)' | (while read f; do g=gcc-19990103/include/${f##gcc/include/}; if [ ! -r $g -a ! -L $g ]; then echo Missing $g; else if [ -f $f ]; then if ! $(cmp -s $f $g); then echo; echo --$f--; diff $f $g; fi; fi; fi; done)

--gcc/include/sys/param.h--
243a244
> 		#undef NULL

--gcc/include/rpc/auth.h--
102a103,106
> 
> #ifdef __cplusplus
> 		void	(*ah_nextverf)(...);
> #else
103a108,112
> #endif
> 
> #ifdef __cplusplus
> 		int	(*ah_marshal)(...);	/* nextverf & serialize */
> #else
104a114,118
> #endif
> 
> #ifdef __cplusplus
> 		int	(*ah_validate)(...);	/* validate varifier */
> #else
105a120,124
> #endif
> 
> #ifdef __cplusplus
> 		int	(*ah_refresh)(...);	/* refresh credentials */
> #else
106a126,130
> #endif
> 
> #ifdef __cplusplus
> 		void	(*ah_destroy)(...);	/* destroy this structure */
> #else
107a132
> #endif

--gcc/include/rpc/svc.h--
83a84,87
> 
> #ifdef __cplusplus
> 	    bool_t	(*xp_recv)(...);	 /* receive incomming requests */
> #else
84a89,93
> #endif
> 
> #ifdef __cplusplus
> 	    enum xprt_stat (*xp_stat)(...); /* get transport status */
> #else
85a95,99
> #endif
> 
> #ifdef __cplusplus
> 	    bool_t	(*xp_getargs)(...); /* get arguments */
> #else
86a101,105
> #endif
> 
> #ifdef __cplusplus
> 	    bool_t	(*xp_reply)(...);	 /* send reply */
> #else
87a107,111
> #endif
> 
> #ifdef __cplusplus
> 	    bool_t	(*xp_freeargs)(...);/* free mem allocated for args */
> #else
88a113,117
> #endif
> 
> #ifdef __cplusplus
> 	    void	(*xp_destroy)(...); /* destroy this struct */
> #else
89a119
> #endif

--gcc/include/idl/c/idl_base.h--
504c504
< #undef NULL
---
> 		#undef NULL

--gcc/include/root/opt/CC/include/codelibs/hashtable.h--
2,3c2,3
< // $Header: hashtable.g,v 1.33 94/07/12 16:08:38 hmgr Exp $
< //--------------------------------------------------------------------*/
---
> /* $Header: hashtable.g,v 1.33 94/07/12 16:08:38 hmgr Exp $*/
> /*--------------------------------------------------------------------*/*/

--gcc/include/math.h--
116a117,122
> #ifdef __cplusplus
> #undef exception
> #endif
> #ifdef __cplusplus
> #define exception __math_exception
> #endif
333a340,346
> struct exception;
> #ifdef __cplusplus
> #undef exception
> #endif
> #ifdef __cplusplus
> #define exception __math_exception
> #endif
337d349
< struct exception;
356a369,374
> #ifdef __cplusplus
> #undef exception
> #endif
> #ifdef __cplusplus
> #define exception __math_exception
> #endif

--gcc/include/mon.h--
22c22
< #undef NULL
---
> 		#undef NULL

--gcc/include/stdlib.h--
37a38,39
> #ifndef _GCC_SIZE_T
> #define _GCC_SIZE_T
38a41
> #endif
116,118c119,121
<      extern char *calloc();
<      extern char *malloc();
<      extern char *realloc();
---
>      extern void *	calloc();
>      extern void *	malloc();
>      extern void *	realloc();

-- 
Marc Girod                Hiomo 5/1          Voice:  +358-9-511 23746
Nokia Telecommunications  P.O. Box 320       Mobile: +358-40-569 7954
NWS/NMS/NMS for Data      00045 NOKIA Group  Fax:    +358-9-511 23580
                          Finland            marc.girod@ntc.nokia.com

duplicate fixinc files

[Marc Girod]

Hi!

I have double copies of some fixinc-* files, in gcc and gcc/fixinc
directories. At least:

fixinc.dgux
fixinc.irix
fixinc.ptx
fixinc.sco
fixinc.svr4
fixinc.winnt
fixinc.wrap

Is it an artifact of my not using cvs correctly, or the mark that the
ones have just been forgotten there?

Best Regards!
Marc

-- 
Marc Girod                Hiomo 5/1          Voice:  +358-9-511 23746
Nokia Telecommunications  P.O. Box 320       Mobile: +358-40-569 7954
NWS/NMS/NMS for Data      00045 NOKIA Group  Fax:    +358-9-511 23580
                          Finland            marc.girod@ntc.nokia.com

Re: duplicate fixinc files

[Marc Girod]

>>>>> "JL" == Jeffrey A Law <law@hurl.cygnus.com> writes:

>> And superficially it looks the same (no time to be more precise now).
>> I'll check deeper later. Tell me if there is something specific you
>> want me to look at.
JL> Excellent!

And more precisely (gcc-19990103 contains the result of the snapshot
configuration, gcc the one of the fixincl-branch):

egcs> find gcc/include | grep -vE '(gcc/include$)' | (while read f; do g=gcc-19990103/include/${f##gcc/include/}; if [ ! -r $g -a ! -L $g ]; then echo Missing $g; else if [ -f $f ]; then if ! $(cmp -s $f $g); then echo Different $f; fi; fi; fi; done)
Different gcc/include/sys/param.h
Different gcc/include/rpc/auth.h
Different gcc/include/rpc/svc.h
Different gcc/include/idl/c/idl_base.h
Different gcc/include/root/opt/CC/include/codelibs/hashtable.h
Different gcc/include/math.h
Different gcc/include/mon.h
Different gcc/include/stdlib.h
egcs> 

Do you want the diffs?

Best Regards!
Marc

-- 
Marc Girod                Hiomo 5/1          Voice:  +358-9-511 23746
Nokia Telecommunications  P.O. Box 320       Mobile: +358-40-569 7954
NWS/NMS/NMS for Data      00045 NOKIA Group  Fax:    +358-9-511 23580
                          Finland            marc.girod@ntc.nokia.com

Re: duplicate fixinc files

[Bruce Korb]

Marc Girod wrote:
> I have double copies of some fixinc-* files, in gcc and gcc/fixinc
> directories. At least:
> 
> fixinc.dgux
> fixinc.irix
> fixinc.ptx
> fixinc.sco
> fixinc.svr4
> fixinc.winnt
> fixinc.wrap
> 
> Is it an artifact of my not using cvs correctly, or the mark that the
> ones have just been forgotten there?

Almost forgotten ;-)  Actually, they are very slightly different
and someday really soon now the files gcc/fixinc.* and gcc/fixincludes
will go away and the code in the gcc/fixinc/ directory will be
activated.  In fact, it should happen this week.  Part of the trouble
is that the three people playing with it have day jobs to do.
The other part is that on one antiquated obscure system there
is an inexplicable failure mode of the new stuff.  We can force
a shell script replacement on those machines, but it leaves room
for some doubt in some people's minds.  You are more than welcome
to play with it on your box to give us a little more of a confidence
base (if you are using a non-GNU-libc system):

	cvs get -rfixincl-branch egcs/gcc

then verify that the ${builddir}/gcc/include is essentially
the same as the non-fixincl-branch version.

The reason for the non-GNU-libc system is that the
fixincludes stuff is a noop on GNU-libc systems.

Re: duplicate fixinc files

[Bruce Korb]

Marc Girod wrote:
> egcs> find gcc/include | grep -vE '(gcc/include$)' | (while read f; do g=gcc-19990103/include/${f##gcc/include/};
 if [ ! -r $g -a ! -L $g ]; then echo Missing $g; 
else if [ -f $f ]; then if ! $(cmp -s $f $g); 
then echo Different $f; fi; fi; fi; done)

> Different gcc/include/sys/param.h
> Different gcc/include/rpc/auth.h
> Different gcc/include/rpc/svc.h
> Different gcc/include/idl/c/idl_base.h
> Different gcc/include/root/opt/CC/include/codelibs/hashtable.h
> Different gcc/include/math.h
> Different gcc/include/mon.h
> Different gcc/include/stdlib.h


Hi Marc, et al.,

I got the copies of the unmodified files and ran several tests.
This test is with the current fixincludes script from the EGCS
repository, with "TARGET_MACHINE" set to "hppa2.0-hp-hpux10.20"

> > sh ./fixincludes $PWD/include3 $PWD/INC                                      
> Building fixed headers in /export/gnu/cygnus/egcs/=build/gcc/include3
> Finding directories and links to directories
>  Searching /export/gnu/cygnus/egcs/=build/gcc/INC/.
> All directories (including links to directories):
> ./usr ./usr/include ./usr/include/sys ./usr/include/rpc ./opt ./opt/CC \
>  ./opt/CC/include ./opt/CC/include/codelibs
> Making symbolic directory links
> Fixing directory =build/gcc/INC into =build/gcc/include3
> Checking header files
> Fixed ./usr/include/sys/param.h
> Fixed ./opt/CC/include/codelibs/hashtable.h
> Cleaning up DONE files.
> Removing unneeded directories:
> ellen:/export/gnu/cygnus/egcs/=build/gcc
> > env | egrep TARGET
> TARGET_MACHINE=hppa2.0-hp-hpux10.20

I patched in my fixincludes so that the other files were not
processed correctly.  More later (see below :).

Notice that only "sys/param.h" and ".../hashtable.h" were fixed.
My new program did not alter "sys/param.h" because it
detected the presence of "#ifndef NULL" in the file already.

All versions altered the ".../hashtable.h", but fixincludes
broke it.  All versions fixed this:

	28c28
	< #if defined sun || defined __sun
	---
	> #if defined __sun__ || defined __sun

additionally, fixincludes did this:

	*** INC/opt/CC/include/codelibs/hashtable.h     Wed May 14 16:24:41 1997
	--- include3/opt/CC/include/codelibs/hashtable.h        Wed Jan 13 11:48:24
1999
	***************
	*** 1,6 ****
	  /*----------------------------------------------------------------------
	! // $Header: hashtable.g,v 1.33 94/07/12 16:08:38 hmgr Exp $
	! //--------------------------------------------------------------------*/
  
	  #ifndef __HASHTABLE_H_
	  #define __HASHTABLE_H_
	--- 1,6 ----
	  /*----------------------------------------------------------------------
	! /* $Header: hashtable.g,v 1.33 94/07/12 16:08:38 hmgr Exp $*/
	! /*--------------------------------------------------------------------*/*/
  
	  #ifndef __HASHTABLE_H_
	  #define __HASHTABLE_H_

which is broken.  None of the rpc files were fixed.  This is what
"auth.h" contains for the AUTH struct (as an example):

    typedef struct __auth {
        struct  opaque_auth     ah_cred;
        struct  opaque_auth     ah_verf;
        union   des_block       ah_key;
        struct auth_ops {
     #ifdef _PROTOTYPES
                void    (*ah_nextverf)(struct __auth *);
                int     (*ah_marshal)(struct __auth *, XDR *);
                /* nextverf & serialize */
                int     (*ah_validate)(struct __auth *,
                                        struct opaque_auth *);
                /* validate varifier */
                int     (*ah_refresh)(struct __auth *);
                /* refresh credentials */
                void    (*ah_destroy)(struct __auth *);
                /* destroy this structure */
     #else
                void    (*ah_nextverf)();
                int     (*ah_marshal)();        /* nextverf & serialize */
                int     (*ah_validate)();       /* validate varifier */
                int     (*ah_refresh)();        /* refresh credentials */
                void    (*ah_destroy)();        /* destroy this structure */
     #endif
        } *ah_ops;
        caddr_t ah_private;
     } AUTH;

The current script detected the fact that there were procedure
pointers declared that did have prototypes, so the patch was not applied.

I started this note yesterday and I still have more analysis to do.
I am sending this now, incomplete.

Thanks for your help.

Regards,
	Bruce

Re: duplicate fixinc files

[Jeffrey A Law]

  In message < 1yg19gye7d.fsf@sinkku.ntc.nokia.com >you write:
  > egcs> find gcc/include | grep -vE '(gcc/include$)' | (while read f; do g=gc
  > c-19990103/include/${f##gcc/include/}; if [ ! -r $g -a ! -L $g ]; then echo
  >  Missing $g; else if [ -f $f ]; then if ! $(cmp -s $f $g); then echo Differ
  > ent $f; fi; fi; fi; done)
  > Different gcc/include/sys/param.h
  > Different gcc/include/rpc/auth.h
  > Different gcc/include/rpc/svc.h
  > Different gcc/include/idl/c/idl_base.h
  > Different gcc/include/root/opt/CC/include/codelibs/hashtable.h
  > Different gcc/include/math.h
  > Different gcc/include/mon.h
  > Different gcc/include/stdlib.h
  > egcs> 
  > 
  > Do you want the diffs?
Yes please.

Thanks,
jeff