| 1 |
> Do you think the problem is with the new hardened gcc specs that |
| 2 |
> automatically adds -fPIC if no -fPIC is found? |
| 3 |
|
| 4 |
no, that's ok as this .o file is loaded dynamically, i.e. it had |
| 5 |
better be PIC (ok, doesn't really matter with the elfloader but |
| 6 |
it does with the dlloader). i just pointed out that this -fPIC |
| 7 |
enforcement ends up creating the reference to that symbol. |
| 8 |
|
| 9 |
> This flipflop logic may be the cause for the .a library getting misbuilt |
| 10 |
> as PIC with the gcc internal preparation function showing up in the |
| 11 |
> object files... |
| 12 |
|
| 13 |
well, one quick way of fixing this problem is to not build X with |
| 14 |
enforced -fPIC then this symbol won't show up either, everyone will |
| 15 |
be happy (as far as the elfloader is concerned but then it has no |
| 16 |
business with PaX ;-). for the future (dlloader) you do want to build |
| 17 |
libbitmap (and probably others) as PIC otherwise you'll get nice |
| 18 |
textrelocs (as i do in my libbitmap.so). and in any case someone will |
| 19 |
have to open a bug with the X guys and tell them about this hidden |
| 20 |
symbol mishandling. |
| 21 |
|
| 22 |
|
| 23 |
-- |
| 24 |
gentoo-hardened@g.o mailing list |
Archives