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]
Other format: [Raw text]

Re: update dwarf2 asm unwind info [hppa64-*-* failures]

John David Anglin wrote: 
There is no problem with symbols in the same section.  For symbols
in different sections, this can be a problem if they lie in different
quadrants (e.g., text and data).  This is why we don't use a pc-relative
relocation for the personality function.  This has to be a plabel
or an indirect to a plabel.

Surely the real reason you don't use a pc-relative relocation for the personality function is dynamic linking.

Anyway, the unwind info that gcc itself emits uses pc-relative relocations in the same locations we want the assembler to do so. It's just that gcc emits funny looking code like

.dword .LFB0+8-$PIC_pcrel$0

which the assembler special-cases instead of

.dword .LFB0-.

which is what the internals of gas assume works when DIFF_EXPR_OK is defined.


r~

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