This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


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

Re: [PATCH] Handle weak symbols


At 07:55 08.05.2001, Mark Mitchell wrote:
> >>>>> "Andreas" == Andreas Jaeger <aj@suse.de> writes:
>
>     Andreas> Are Franz' patches broken or not?  I'd rather see this
>     Andreas> fixed now on the mainline.
>
>Broken is a strong term.
>
>They are not sufficiently perfect at this point.
>
>I think the right way to do this is, in broad strokes:
>
>   1. The front-end sets DECL_WEAK when it creates a declaration
>      for which `#pragma weak' has been seen.  This is easy,
>      but we should still use a data structure smarter than a
>      linked list to store the list of `#pragma weak' names.

I don't think it's worth to do that for a seldom used #pragma weak.

>   2. Similarly, if the declaration already exists, it is marked
>      DECL_WEAK when a `#pragma weak' is seen.
>
>      This step is non-trivial, but doable.  We could perhaps also make
>      it illegal to use `#pragma weak' after the function has already
>      been declared, which would obviate this step.  After all,
>      GNU code will use __attribute__, not #pragma.

That was my intention for a followup patch.


>   2. make_decl_rtl should set SYMBOL_REF_WEAK from DECL_WEAK, as
>      per Franz's patch.
>
>   3. rtx_addr_can_trap_p should check SYMBOL_REF_WEAK as per
>      Franz's patch.

As you seem to agree with my way to implement these 2 points, can only that 
go in? These 2 are enough to fix the glibc problem for now. We can work on 
the rest of the patch in the mainline.

Franz.


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