Re: New CVS snapshot 20000213

From: Harry E. Clarke (hec@heclarke.freeserve.co.uk)
Date: Mon Feb 21 2000 - 13:02:05 PST

  • Next message: Randolph Bentson: "Preview functionality"

    I guess I must of have been realy unlucky then for my K&R code to behave
    differently on two different systems. Its too many years ago to remember
    the reason, but it was something along the lines of the way different
    compilers made asumptions.

    Of course in K&R code you ignore parameter checking all together.
    Just one less set of warnings to have to worry about.

    Have fun
    Harry

    In message <CA25688B.0082FBB8.00@d73mta01.au.ibm.com>, msitkows@au1.ibm.com wri
    tes:
    >
    >
    >I beg to differ on this one since, in my experience, the one thing ANSI C
    >is not, is 'portable'.
    >After 15 years of writing C on almost every platform known to man (and some
    >which aren't...) I can almost guarantee that code which strictly adheres to
    >K&R rules will compile cleanly on any platform (given a few system call
    >differences, like vfork() etc).
    >However, I have yet to find two ANSI C compilers which agree on all points
    >of the so-called standard. You'll find they cautiously refer to 'levels of
    >compliance'.
    >I remember 'translating' to ANSI some quite ordinary X11/Motif K&R code
    >which ran very nicely on AIX, SGI and Sun. It finished up being peppered
    >with hundreds of '#ifdef AIX' '#ifdef SGI' statements, and inherited some
    >interesting bugs.
    >If you only support one compiler, like gcc, then you might be able to port
    >the same code to most platforms. However, first read the platform specific
    >notes that come with gcc - it doesn't *quite* work the same on every
    >box....
    >
    >Best regards,
    >
    >Mark
    >------------------------------------------------------------------------
    > Mark Sitkowski
    > Design Simulation Systems Ltd
    > 14 Loddon Street
    > Box Hill North, Victoria 3129
    >------------------------------------------------------------------------
    > phone: (613-9) 897-1375 fax: (613-9) 897-1376
    > Mobile: 0413-890-777 E-mail: marks@iaccess.com.au
    > Web: www.angelfire.com/de/designsim
    > www.asiafind.com/sites/designsim
    >------------------------------------------------------------------------
    >
    >There has been discussion recently about adding/removing compiler warnings.
    >If you want to write good clean code that is portable, then use ANSI C
    >compiler (not allowing tradtional C), and remove all warnings. The
    >following
    >flags should be used.
    >
    > -ansi -pedantic -Wall
    >
    >Adding the flag "-Wpointer-arith", is probably a good idea, or at least
    >a very good reason why the warnings are OK.
    >
    >As the ansi and pedantic flags are not currently used, they should be
    >introduced for the next release, after the release on Sunday.
    >
    >Harry
    >
    >
    >In message <38AE7D8F.9A0334B8@dial.pipex.com>, John Vickers writes:
    ><snip>
    >>> Therefore I suggest to remove the flag -Wpointer-arith from the final
    >>> release.
    >>
    >>Or maybe remove all the warning flags, even. I get huge numbers of
    >>warnings
    >>building Sane with "-W -Wall". But for development releases,
    >>we should really be trying to get rid of the causes of the warnings.
    >>
    >
    >--
    >Source code, list archive, and docs: http://www.mostang.com/sane/
    >To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
    >
    >
    >
    >
    >--
    >Source code, list archive, and docs: http://www.mostang.com/sane/
    >To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
    >

    --
    Source code, list archive, and docs: http://www.mostang.com/sane/
    To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
    



    This archive was generated by hypermail 2b29 : Mon Feb 21 2000 - 13:00:56 PST