| 1 |
On Wed, 2009-07-22 at 10:22 +0200, Fabian Groffen wrote: |
| 2 |
> On 22-07-2009 10:10:32 +0200, Markus Duft wrote: |
| 3 |
[snip] |
| 4 |
> |
| 5 |
> I think the following will get us up to speed quickly: |
| 6 |
> - rename the current script to "readpecoff" or something |
| 7 |
> - ship it with portage for the time being |
| 8 |
|
| 9 |
ok. i updated the patch -> attached. |
| 10 |
|
| 11 |
> |
| 12 |
> Then I'd like it when we could get a scanpecoff (like scanmacho and |
| 13 |
> scanelf, the C-programs from pax-utils) that does the job, replacing |
| 14 |
> your huge bash-script. |
| 15 |
|
| 16 |
puh. i could try to implement such a thing, but to be correct i would |
| 17 |
need one for interix, and one for windows - that the "readpecoff" script |
| 18 |
can handle both is more a hack than a solution ;) |
| 19 |
|
| 20 |
OTOH - why not..? it feels like a good idea to have such a utility. it |
| 21 |
could detect wether the file is interix/winnt and handle it correctly - |
| 22 |
maybe i could also lookup the interix shared library section format from |
| 23 |
the binutils patch, and implement it there - so that scanpecoff is |
| 24 |
faster than parsing objdump output all the time... |
| 25 |
|
| 26 |
> |
| 27 |
> > > At the moment I'm trying to open a docx file from Microsoft that should |
| 28 |
> > > be the specification of PECOFF. |
| 29 |
> > |
| 30 |
> > uh... you won't have fun with that document - i know it by rote ;) it |
| 31 |
> > does _not_ specify how both interix and winnt handle runpaths, needed |
| 32 |
> > shared libs, etc. |
| 33 |
> |
| 34 |
> I can't even open it. OpenOffice die hard aborts on it (Signal 6). |
| 35 |
|
| 36 |
hahaha ... |
| 37 |
|
| 38 |
> |
| 39 |
> > how interix handles those things is completely hidden somewhere in the |
| 40 |
> > interix loader/binary format/binutils patches at M$. (However the |
| 41 |
> > current binutils 2.17 patches are open, so i could have a look at it - |
| 42 |
> > the patch is about 300.000 lines long ;*( ...) |
| 43 |
> |
| 44 |
> I once split that patch into our binutils ebuild for 2.17, maybe it's |
| 45 |
> useful? |
| 46 |
|
| 47 |
ah, you did that - i was wondering why the hell the binutils have such a |
| 48 |
big patch - i could not remember doing it ;) however objdump from our |
| 49 |
binutils is still unable to dump the "Dynamic (.so) Section" along with |
| 50 |
the other private PE headers :(. |
| 51 |
|
| 52 |
i will not investigate on this further if i'm able to extract the |
| 53 |
information in the future shiny new scanpecoff from pax-utils. if i'm |
| 54 |
unable to get scanpecoff to read interix binaries by itself, i will try |
| 55 |
to fix this, so that we can at least use our own objdump instead of the |
| 56 |
system one... |
| 57 |
|
| 58 |
Cheers, Markus |
| 59 |
|
| 60 |
> |
| 61 |
> > The format used by winnt to model shared library features (rpath, |
| 62 |
> > needed, etc.) is defined by parity (native win32 does not support those |
| 63 |
> > out of the box...) - it builds upon the PE/COFF specs, but of course it |
| 64 |
> > does it's own things :) if you want to know what it does exactly -> ask |
| 65 |
> > me, as i wrote it ;) |
| 66 |
> |
| 67 |
> |
Archives