This is the mail archive of the binutils@sources.redhat.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: PE bug: bad DataDirectory.


If I remember the code correctly, the initialization of entries 1 and 12
can only be (reasonably) done where it is (or "nearby") because the
information
needed to do it isn't available in ...swap_aouthdr_out.  (It was the
case
at the time I last looked; things could have changed.)

The entries that are initialized in swap_aouthdr_out could (possibly) be
initialized
in ...final_link_postscript(?).

It appears that removing the memset completely introduces no "junk" into
the
Data Directory, so a selective memset may not be needed.  (Or maybe the
memset
belongs in ...final_link_postscript.)

Based upon these and other comments, my current "feel" for the right
answer is
to do everything in ...final_link_postscript, but that presumes that the
information
to do so is available at that time.  Anyone know that for sure (either
way) before
we go digging in the code?

Donn

-----Original Message-----
From: Nick Clifton [mailto:nickc@redhat.com] 
Sent: Friday, August 16, 2002 9:14 AM
To: Donn Terry
Cc: binutils@sources.redhat.com
Subject: Re: PE bug: bad DataDirectory.


Hi Donn,

> Scenario: ldwrite calls _bfd_coff_final_link which calls 
> _bfd_pei_final_link_postscript. _bfd_pei_final_link_postscript fills 
> in the import table entries for the .idata section.
> 
> *Subsequently* main calls bfd_close calls coff_write_object_contents 
> calls _bfd_pei_swap_aouthdr_out.  The memset in 
> _bfd_pei_swap_aouthdr_out zeros out the work that 
> _bfd_pei_final_link_postscript did.  Oops.

It appears that _bfd_pei_final_link_postscript only fills in entries 1 &
12 of the DataDictionary, and that apart from the memset
_bfd_pei_swap_aouthdr_out only initialises entries 0, 1, 2, 3 & 5.

I am not sure if ..aouthdr_out's initialisation of entry 1 is correct,
but assuming that it is, then it would seem that the memset should be
restricted to entries 4, 6, 7, 8, 9, 10, 11, 13, 14, 15 & 16.   Either
that or code to initialise entry 12 should be added to ...aouthdr_out().

What do you think ?

Cheers
        Nick



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]