Hi, I convinced myself that this needs to be fixed before the release. While I am sympathetic to Vladimirs argument this seems serious enough to me to call it a showstopper. Fernando Cacciola wrote:
Hi Richard,
I' glad you found this regression now, _before_ the release.
FWIW I agree that that 1.34 should ship with an non-breaking implementation of none. As you know the intention was to fix a long standing problem with precompiled headers, but by no means that fix can existing code.
I will contact the Release Manager (Thomas) asking for permission to revert the fix that caused the regresion.
I will patch it by physically reverting the problematic fix, from CVS history, rather than patching the code one more time, to make sure there are no further mistakes.
Please go ahead. It's your call whether to simply revert or accept Richard's patch. The way I look at it is that the patch has higher risk but a nice testcase, that makes it even to me. Thomas -- Thomas Witt witt@acm.org