Re: Warning platform maintainers

From: Mike Nordell (tamlin@algonet.se)
Date: Mon Dec 02 2002 - 18:45:49 EST

  • Next message: Mike Nordell: "Re: commit: abi: ugly hack"

    Hubert Figuiere wrote:

    > In the last commit I fixed a double free of the popt context in the
    > Cocoa code. While reviewing, I found out that it was caused by tamlin's
    > leak fixes in the Win32 code.

    Ouch. Sorry Hub, obviously that wasn't my intention. But wouldn't it make
    more sense to let the arg class handle both the resource aquisituion and
    release instead of repeating the same code in every platform implementation,
    opening up for mistakes and misunderstandings like this?

    Not only would it enhance encapsulation and making sure stuff like this
    don't happen, it would also force the arg class to provide a functional
    interface, opening the door for using whatever argument parsing lib one
    desires, without changing any code in the user/client code using the arg
    class, with only one single place to change - ap_Args.cpp.

    If it doesn't make sense: why? What's the benefit of global static data that
    anyone can probe/set/reset/screw up compared to a class encapsulating the
    resource aquisition/release and its interface providing the necessary
    operations?

    (hmmm, this reminds me of an old AW+libpng "incident"... :-> )

    /Mike



    This archive was generated by hypermail 2.1.4 : Mon Dec 02 2002 - 18:48:58 EST