Re: [PATCH] pe/coff - add support for base64 encoded long section names

[Tristan Gingold <tgingold@free.fr>]

On 22/05/2023 08:36, Jan Beulich wrote:
[...]
> I wonder whether this wouldn't better be handled either by returning
> the full 36-bit value and then letting the caller decide (it wants to
> check against string table size anyway, not just 2³²), or by checking
> that the first digit is only 'A'...'D'. The latter would of course
> make the helper function less generic (if any further use appeared).

I prefer to keep the overflow check and also to keep the function generic.
Using a 64b type might be tricky as it is easy to have the value 
silently truncated due to the use of a 32b type.

> I further wonder what LLVM's motivation is to zero-pad (i.e.
> 'A'-prefix) the encoded values, rather than - like pre-existing
> section name representation - nul-padding at the end. At the very
> least I'm inclined to suggest that we also support that obvious (I
> think) alternative here.

I don't see the point. The nul-padding is not used by anyone, and 
therefore cannot even be tested.  Why creating dead code ?

I will submit a v2 of the patch.

Tristan.