Re: [patch]: make idata$4 and .idata$5 more compliant to coff spec

["Kai Tietz" <ktietz70 at googlemail dot com>]

Hi Nick,

2009/1/2 Nick Clifton <nickc@redhat.com>:
> Hi Kai,
>
>> this patch cleans up an unnecessary spec violation for coff. For each
>> import library linked into an executable there are leading zero
>> elements added to idata$4 and idata$5. By this patch those elements
>> aren't wrote anymore. I tested it for arm-mingwce, x86_64-pc-mingw32
>> and for i686-pc-mingw32 without seeing any regressions.
>
> I am not so happy about this - those NUL entries appear to be there for a
> reason.  What is the spec that is being violated ?
>
> Cheers
>  Nick
>
> PS.  When you checked for regressions, did you also check that dll's built
> with the patched toolchain still work in the real world ?

Yes, I verified this. Btw binutils now builds dll's the same way as VS does it.
I am also not verify happy about this change, because it seems to be
here for decades. I came to this by trying to add IAT directory entry
support and found that windows resists when I use the .idata$5 section
(which is in fact IAT) has a leading zero.

The spec you can get by ms and is the coff v8.1 spec. If you like I
can sent it too you per e-mail. For this news group it would be a bit
too big, I guess.

Cheers,
Kai


-- 
|  (\_/) This is Bunny. Copy and paste
| (='.'=) Bunny into your signature to help
| (")_(") him gain world domination