Re: [aida_s@mx12.freecom.ne.jp: A serious bug of "ld --enable-auto-import"]

[Charles Wilson <cwilson@ece.gatech.edu>]

DJ Delorie wrote:

>>pe_find_data_imports:_hwstr1
>>->__head_cyghwstr_dll
>>arelent: _hwstr1@0xd: add=0  <<<< this should be "add=12", right?
>>
> 
> Not always.  Sometimes the addend is in the relent, and sometimes the
> addend is in the data stream.  It depends on the target format.
> 


Well, in *this* case we're talking about pe-386....

> 
>>Unfortunately, that's as far as I can go.  I don't know where the 
>>arelent structure gets filled with the approprate data from the .o, so I 
>>can't track down WHY the "12" isn't going into (arelent)->addend.
>>
> 
> Gas puts it there.  Probably in src/gas/config/tc-i386.c's
> tc_gen_reloc, but other places convert it from a bfd reloc to an
> i386-specific reloc.


Hmm...but the original bug report said (and I verified) that it works 
fine if you add the appropriate __declspec(dll??port) modifiers. 
Somehow, this is tied to the autoimport additions -- but none of that 
touched gas.

Anyway, I'm confused.  The .o file already contains the "12" (actually, 
0c 00 00 00 in on-disk byte order), so gas has already done its work -- 
properly.  The problem occurs during the linking step -- ld is ignoring 
the offset stored within the opcode and is blindly stuffing in the 
relocation address without adding the offset.

That doesn't involve gas at all, does it?--gas is done, at this point. 
As *ld* runs, it has to fill arelent structures in order to actually 
*perform* the relocation, right?  *That's* the part that is not 
happening correctly.

--Chuck