[Bug ld/5526] eh support is broken

[rsandifo at nildram dot co dot uk]

------- Additional Comments From rsandifo at nildram dot co dot uk  2008-01-02 
15:20 -------
> Regarding the frame table error, I have collected all the init and
> libraries used in linking the libstdc++-v3 abi_check test.  The is
> a script yyy.sh that produces the following output on hppa-linux:

Thanks.  I see the problem.  The information we use to compare
personality routines is not available at GC-marking time.
I'll attach a patch in sec.

> The other error is a runtime error, so I don't see how it can be debugged
> other than in situ.  I was hoping that you might see why the encoding that
> GCC uses doesn't work with your change.  It does rely on there being a
> fixed relationship between text and data in shared objects.

Unfortunately, it's not particularly obvious from the encoding alone.
The marking code doesn't really make any assumptions about the encoding
type used, so if we can parse the .eh_frame at all, we should mark the
associated section correctly.  And regarding the "if": the patch doesn't
really change _how_ the .eh_frames are parsed, just when.

I suppose the bug that caused the abi_check problem might have caused
this one too.  Definitely worth a go.

Richard


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED


http://sourceware.org/bugzilla/show_bug.cgi?id=5526

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.