Am Sonntag, den 21.09.2008, 11:25 +0200 schrieb Alexandre Julliard:
Michael Karcher wine@mkarcher.dialup.fu-berlin.de writes:
OK. Thanks for the pointer. I have a configure test ready that checks whether the global -fno-builtin is really needed., and uses separate flags otherwise. I am planning to submit it in series with a second patch that checks for whether -fno-builtin-scanf is needed (check for warning on "%P").
I don't think there's any reason to make it that complex. You can just use -fno-builtin in all cases.
Sure I can, this was approach b I suggested. But that may be harmful to performance, as it forbids gcc to use any knowledge about the standard library (includes inlining of memcpy with small constant sizes or built-in pureness annotations).
But the issue that started the discussion was: What to do about the scanf warning? On a system that does not have wchar builtins in gcc, currently all built-in functions are enabled. Even on these systems, we should disable the built-in scanf to prevent the warning in the scanf test. This need has long been unnoticed, as wchar ctype functions are built-in on our main development platform.
Regards, Michael Karcher