[PATCH 2/3] [binutils/readelf] Add fixme in display_debug_str_offsets

[Tom de Vries]

On 5/16/24 02:14, Alan Modra wrote:
> On Wed, May 15, 2024 at 04:18:58PM +0200, Tom de Vries wrote:
>> On 5/15/24 15:02, Alan Modra wrote:
>>> Sorry, that doesn't look like a realistic testcase to me.  An offset
>>> of 0xffffffff in 32-bit dwarf?
>>>
>>
>> I'm not claiming that this is a realistic test-case.
>>
>> That's what I meant in the commit message with "rare and odd".
> 
> It isn't "rare and odd".  Impossible to occur in any but fuzzed object
> files is more like it.  I definitely do not want to see an assertion
> (resulting in bug reports) that can only occur with fuzzed objects,
> and where failure of the assumption does not cause any bad behaviour.
> 
> By "bad behaviour" I mean
> - segfaults and similar,
> - buffer overflow,
> - infinite looping,
> - and maybe OOM.
> 

I'm not proposing to add an assertion.

As mentioned, the patch containing the assert is a demonstrator patch.

The purpose of the proposed patch is to fixate and document the current 
state of things (which needs fixing).  If that's not acceptable, I can 
open a PR to document it.

Thanks,
- Tom