The following problem was discovered using a cross tool chain --
x86 linux x powerpc eabi. However, I have reproduced it using
an x86 linux native toolchain. I do not *believe* that it is
dependent upon any of host triplet, target triplet, build triplet.
Suppose you have a linker script which explicitly assigns
the .data sections from some selected files to .nodata and
*LATER* wildcards .data to go into .data.
You would expect that the early explicit assignments would
be honored. They, at least sometimes, aren't.
I have a greatly stripped down example using 3 object files
(the original involved over 700 object files). It can probably
be reduced further.
linker-script:
--------------
SECTIONS {
cs 0x2e000 : {
"foo.o"(.text .rodata .rodata.* .sdata2)
"bar.o"(.text .rodata .rodata.* .sdata2)
} =0x00000000
.data 0x890000 : {
"fubar.o"(.data .sdata)
} =0x00000000
.bss 0xd10000 : {
"fubar.o"(.bss .sbss)
} =0x00000000
.nodata 0xffffff00 : {
"foo.o"(.data .sdata)
"bar.o"(.data .sdata)
} =0x00000000
.nobss 0xffffff80 : {
"foo.o"(.bss .sbss)
"bar.o"(.bss .sbss)
} =0x00000000
$newlib$ (ADDR(cs))+SIZEOF(cs) : {
. = ALIGN(4);
*(.text .rodata .rodata.cst8 .rodata.cst4 .eh_frame .got)
. = ALIGN(16);
} =0x00000000
.data 0x890000 : { *(.data .sdata) }
.debug 0 : { *(.debug) }
.debug_srcinfo 0 : { *(.debug_srcinfo) }
.debug_aranges 0 : { *(.debug_aranges) }
.debug_pubnames 0 : { *(.debug_pubnames) }
.debug_sfnames 0 : { *(.debug_sfnames) }
.line 0 : { *(.line) }
.stab 0 : { *(.stab) }
.stabstr 0 : { *(.stabstr) }
/DISCARD/ : { *(.comment .alias .fixup .data.rel) }
}
bar.c:
------
int bar_data = 42;
int
bar_function()
{
return 42;
}
foo.c:
------
int foo_data = 17;
void
foo_function()
{
}
fubar.c:
--------
int fubar_bss;
int fubar_data = 29;
instructions:
-------------
compile the files (-c) using your favorite gcc
link 'em saying
<some linker> -static -Map bug.map -o bug.elf --script=linker-script
examine using objdump --headers and the file bug.map.
You'll see that
.nodata did not get created; it should have been create with the .data .sdata
contents from foo.o and bar.o. The contents instead went into .data.
--
Summary: explicit input section assignment overriden by *LATER*
wildcard assignment
Product: binutils
Version: 2.16
Status: NEW
Severity: normal
Priority: P2
Component: ld
AssignedTo: unassigned at sources dot redhat dot com
ReportedBy: dtaylor at emc dot com
CC: bug-binutils at gnu dot org
http://sourceware.org/bugzilla/show_bug.cgi?id=3089
------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.
_______________________________________________
bug-binutils mailing list
bug-binutils@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-binutils
