[RFC] Cygwin libstdc++ plan (operator new/delete replacement)

Dave Korn dave.korn.cygwin@googlemail.com
Fri Oct 2 14:29:00 GMT 2009

Corinna Vinschen wrote:
> On Oct  2 01:54, Christopher Faylor wrote:

>> No, this has to stop at some point.  The current workaround is adequate
>> and really should go away soon; maybe even for 1.7.1.  We're not going
>> to stand on our heads for a handful of dlls which might have been built
>> using beta code and might have overridden the cxx_malloc structure.  I
>> made that decision when I added this kludge.

  It's not just a handful of DLLs, it's also every executable that was built
before we repurposed the forkee field.

> ACK.  If the current code for cxx_malloc and friends is right (apart
> from the kludge to get broken DLLs working), the right solution is to
> fix the broken DLLs built against the broken test Cygwin DLL, rather
> than adding more kludges to support them indefinitely.  It would be
> different if this had been official Cygwin releases, but fortunately
> these were just betas.

  Ok, not writing more code suits me!  But I don't see that we can take away
the kludge because it fixes the zeroisation that is done by far more things
than just the handful of DLLs that were redirecting the pointer.


More information about the Cygwin-developers mailing list