snapshot 19990608 libstdc++ check error

Eric Raskin ehr@listworks.com
Wed Jun 9 11:44:00 GMT 1999


Per Jeffrey's request, here are the differences I've found.  First are the things needed by DG/UX:

1)  fix for NULL definitions (made conditional and correct for C/C++) in:
	curses.h  dbm.h  fmtmsg.h  locale.h  loctalias.h  mon.h  stddef.h  stdio.h
	stdlib.h  string.h  time.h  timers.h  sys/_int_unistd.h  sys/param.h
	rpc/types.h  X11/Intrinsic.h  X11/Xlibint.h  X11/Xlocale.h
 
2)  fix for declaration of struct exception in:
	math.h

3)  fix for declaration of struct queue in:
	queue.h

4)  replace _int_varargs.h with DG/UX specific version (allows for 88K and Intel
    to co-exist in same include file based on conditional pre-processing).

5)  Add conditional compilation based on _KERNEL in:
	sys/err.h

The following fixes listed in fixinc.svr4 don't seem to be needed on DG/UX:

1)  #lint(on), #lint(off), #machine(), #system(), #cpu(), machine specified #if's,
    __STDC__ comparisons, etc.  They aren't used in the DG headers.

2)  Fixing size_t, ptrdiff_t, wchar_t:
	sys/types.h  stdlib.h  sys/stdtypes.h  stddef.h  memory.h  unistd.h

3)  Fix getcwd definitions in:
	stdlib.h  unistd.h 

4)  Fix mmap definition in:
	sys/mman.h

5)  Create static or extern definitions for ftw and nftw:
	ftw.h

6)  Don't define bool type in:
	curses.h

7)  Add static void getrnge in:
	regexp.h

8)  Remove #defines of some string functions in:
	string.h

9)  Remove #defines of __i386, __i860, __mips in:
	ieeefp.h

10) Add #define for _SIGACTION_ and fix types of SIG_DFL, SIG_ERR, SIG_IGN, and SIG_HOLD in:
	sys/signal.h

11) Fix declarations of makedev, major, and minor in:
	sys/mkdev.h

12) Fix reference to NMSZ in:
	sys/adv.h (doesn't exist on DG/UX)

13) Fix reference to NC_NPI_RAW in:
	sys/netcspace.h (doesn't exist on DG/UX)

14) Add conditional compilation on _KERNEL to:
	fs/rfs/rf_cache.h (doesn't exist on DG/UX)
	sys/erec.h (doesn't exist on DG/UX)
	sys/char.h (doesn't exist on DG/UX)
	sys/getpages.h (doesn't exist on DG/UX)
	sys/map.h 
	sys/cmn_err.h
	sys/kdebugger.h (doesn't exist on DG/UX)

15) Check __GNUC__ and __GNUG__ in:
	sys/endian.h

16) Remove useless extern keyword from forward struct declarations in:
	sys/stream.h
	sys/strsubr.h (doesn't exist on DG/UX)

17) Put storage class at start of decl to avoid warning in:
	rpc/types.h

18) Convert functions to proto form and fix arg names in:
	sys/stat.h

19) Test for /bin/sony and repairs

20) Put #ifndefs around items that are defined in float.h:
	limits.h

21) Put #ifndefs around HUGE_VAL in:
	math.h

22) fix proto definitions in:
	math.h  floatingpoint.h

23) Replacing sys/byteorder.h (DG/UX headers already optimize using asm() instructions)


So, what's next.  Do I follow Robert's advice and try to convert to fast-fixincludes?

    Eric Raskin
	

> -----Original Message-----
> From: Jeffrey A Law [ mailto:law@cygnus.com ]
> Sent: Wednesday, June 09, 1999 1:15 PM
> To: ehr
> Cc: 'Robert Lipe'; egcs@egcs.cygnus.com
> Subject: Re: snapshot 19990608 libstdc++ check error 
> 
> 
>   In message 
> < 004201beb296$41fd7240$65c8c8c8@ehrpc.listworks.com >you write:
>   > Robert:
>   > 
>   > I use fixinc.dgux -- which I guess hasn't been updated. 
> I'm going to take t
>   > he the fixinc.svr4 to pull out the differences and create 
> a patch for fixin
>   > c.dgux.
> What I would recommend first is noting the differences and 
> telling the list
> what things fixinc.svr4 does that are not done by 
> fixinc.dgux.  That way we
> can quickly say "we need to port X, Y & Z, but not W to fixinc.dgux".
> 
> This is an excellent example of why system specific fixinc 
> files are a bad
> idea and why they should go away long term with everyone, and 
> I mean everyone
> using the same fixinc code.
> 
> jeff
> 
> 
> 
> 



More information about the Gcc mailing list