Re: [PATCH, ARM] Fix PR51643

2011-12-22 Thread Richard Earnshaw
On 22/12/11 17:06, Richard Henderson wrote: > On 12/22/2011 06:18 AM, Richard Earnshaw wrote: >> + && lookup_attribute ("weak", DECL_ATTRIBUTES (decl))) > > DECL_WEAK (decl) > > > r~ > Dur, fixed. R.

Re: [PATCH, ARM] Fix PR51643

2011-12-22 Thread Richard Henderson
On 12/22/2011 06:18 AM, Richard Earnshaw wrote: > + && lookup_attribute ("weak", DECL_ATTRIBUTES (decl))) DECL_WEAK (decl) r~

[PATCH, ARM] Fix PR51643

2011-12-22 Thread Richard Earnshaw
The ARM bare-metal configuration of the EABI specifies that an unresolved weak direct call becomes a NOP rather than faulting; but this is only supported for the BL instruction. GCC mishandled this in some cases by trying to convert such a call into a tail-call, but B instructions are not covered