[ECOS] eCos compile problem with arm-none-eabi- toolchain 2008q1

["Xie Chun-B19413" <Chun dot Xie at freescale dot com>]

Hi Simon,
	We use arm-none-eabi- toolchain 2008q1 release compile eCos. The
compiler option -mcpu=cortex-a8 was used.
	And we met a strange problem. The compiler/linker couldn't
calculate the right offset of the string pointer of function
diag_printf. 
	
	For example,
	
	//some code
	diag_printf("string_1");
	diag_printf("string_2");
	//some other diag_printf was called

	diag_printf("string n");			//this function
string pointer will not point to it's "string_n", but it will return to
somewhere of "sting_1"

	//other diag_printf was called

The following is the dump of the code.

     2397c:       e3040ffc        movw    r0, #20476      ; 0x4ffc
   23980:       e3400005        movt    r0, #5  ; 0x5
   23984:       eb005229        bl      38230 <diag_printf>
//string pointer points(r0) to 0x54ffc
   23988:       e3040014        movw    r0, #16404      ; 0x4014
   2398c:       e3400005        movt    r0, #5  ; 0x5
   23990:       eb005226        bl      38230 <diag_printf>
//string pointer points(r0) to 0x54014, it returns to the origin address
of .rodata,where stores all the strings.
	
//actrually this pointer should point to an address behind 0x54ffc.



Disassembly of section .rodata:
00054008 <__func__.3237-0x18fc>:
   54008:       636c6557        
   5400c:       20656d6f       
   54010:       46206f74        
   54014:       73656572        


And arm-none-eabi- toolchain 2008q3 release could compile this code
correctly.

But we still want to know the reason of this problem. We hope you could
help us.

Best Regards,
Alex


--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss