This is the mail archive of the gdb-testers@sourceware.org mailing list for the GDB 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]

[binutils-gdb] Optimize R_386_GOT32/R_386_GOT32X only if addend is 0


*** TEST RESULTS FOR COMMIT 02e2aef89bb58e8e0cc4390db41c5f775e1b7c3e ***

Author: H.J. Lu <hjl.tools@gmail.com>
Branch: master
Commit: 02e2aef89bb58e8e0cc4390db41c5f775e1b7c3e

Optimize R_386_GOT32/R_386_GOT32X only if addend is 0

Linker can't optimize R_386_GOT32 and R_386_GOT32X relocations if addend
isn't 0.  It isn't valid to convert

movl	foo@GOT+1(%ecx), %eax

to

leal	foo@GOTOFF+1(%ecx), %eax

nor to convert

movq	foo@GOTPCREL+1(%rip), %rax

to

leaq	foo(%rip), %rax

for x86-64.  We should check if addend is 0 before optimizing R_386_GOT32
and R_386_GOT32X relocations.  Testcases are added for i386 and x86-64.

bfd/

	* elf32-i386.c (elf_i386_convert_load): Skip if addend isn't 0.
	(elf_i386_relocate_section): Skip R_386_GOT32X optimization if
	addend isn't 0.

ld/testsuite/

	* ld-i386/i386.exp: Run mov2a, mov2b and mov3.
	* ld-i386/mov2.s: New file.
	* ld-i386/mov2a.d: Likewise.
	* ld-i386/mov2b.d: Likewise.
	* ld-i386/mov3.d: Likewise.
	* ld-i386/mov3.s: Likewise.
	* ld-x86-64/mov2.s: Likewise.
	* ld-x86-64/mov2a.d: Likewise.
	* ld-x86-64/mov2b.d: Likewise.
	* ld-x86-64/mov2c.d: Likewise.
	* ld-x86-64/mov2d.d: Likewise.
	* ld-x86-64/x86-64.exp: Run mov2a, mov2b, mov2c and mov2d.


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