[U-Boot] Bug in TOP860 code with gcc 4.8.1

[Reinhard Meyer]

Hello Jeroen,
> Hello Reinhard,
>
> On 01/14/2014 12:33 PM, Reinhard Meyer wrote:
>> Dear Wolfgang,
>>> Dear Reinhard,
>>>
>>> attempting to build the TOP860 code with a GCC 4.8.1 based tool chain
>>> (say ELDK v5.5 or Yocto 1.5) gives the following errors:
>>>
>>> -> ./MAKEALL TOP860
>>> Configuring for TOP860 board...
>>>     text    data     bss     dec     hex filename
>>>   165471   21020   17316  203807   31c1f ./u-boot
>>> ../common/flash.c: In function 'flash_init':
>>> ../common/flash.c:336:20: warning: iteration 128u invokes undefined 
>>> behavior [-Waggressive-loop-optimizations]
>>>       info->start[i] = (ulong)addr + 0x10000 * i;
>>>                      ^
>>> ../common/flash.c:334:4: note: containing loop
>>>      for (i = 0; i < info->sector_count; i++)
>>>      ^
>>> ...
>>>
>>> Can you please provide a fix - or is this old hardware and the code
>>> should be removed from the U-Boot tree?
>> 1. on first and second glance I cannot see where this (simple!!) loop 
>> might "invoke undefined behaviour". Seems like a compiler/optimizer 
>> bug to me...
>>
>> 2. should not the same issue arise with TOP5200 (using the same 
>> flash.c) ??
>>
>> 3. nevertheless TOP860 can be removed.
>
> It is out of bounds:
>
> include/configs/TOP860.h:#define CONFIG_SYS_MAX_FLASH_SECT    128 /* 
> max number of sectors on one chip    */
> include/configs/TOP5200.h:#define CONFIG_SYS_MAX_FLASH_SECT 256 /* max 
> num of sects on one chip */
>
> Removing will work as well of course ;)
>
> Regards,
> Jeroen
I see. Because of the common code with TOP5200, there is potentially a 
larger flash chip available in the switch, which would cause undefined 
behaviour later. However TOP860 has never seen more than 2MB of flash :) 
so there is no actual danger.

For the next few weeks I do not have the capacity to provide a patch, 
and since the TOP860 U-Boot is frozen to a very old state by the 
customers anyway, it is save to be removed from the current tree.

Best Regards
Reinhard