This is the mail archive of the binutils@sourceware.org mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: equating symbols to undefined


On Thu, Oct 13, 2005 at 09:05:54AM +0200, Jan Beulich wrote:
> >>> "H. J. Lu" <hjl@lucon.org> 12.10.05 18:06:45 >>>
> >On Wed, Oct 12, 2005 at 05:31:01PM +0200, Jan Beulich wrote:
> >> 
> >> I want to be able to have a header file that defines an alias
> memcpy
> >> should resolve to (because there is no way I could tell gcc to not
> call
> >> memcpy when copying large compound objects), which can be included
> by
> >> both consumers and producer of the replacement definition, without
> the
> >> producer (because of inclusion of the header file) providing a
> global
> >> definition for memcpy.
> >> 
> >> Similarly, I want to be able to a header file that defines an
> alias,
> >> say, __divdi3 should resolve to (because there is no way I could
> tell
> >> gcc to not call __divdi3 when dividing 64-bit integers on a 32-bit
> >> target), which can be included by both consumers and producer of
> the
> >> replacement definition, without the producer (because of inclusion
> of
> >> the header file) providing a global definition for __divdi3 (because
> the
> >> policy in the rest of the component - linux kernel - is to not have
> such
> >> a function available).
> >> 
> >
> >Then please show me a testcase where I can do
> >
> ># make
> >
> >and see the failure.
> 
> Here we go: just 'make' results in two calls to copymem (as logged to
> the console), whereas 'make DEFINES=LOCAL' (with an older assembler used
> by the compiler) results in just one call (as intended).
> 

Where is the testcase?


H.J.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]