[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

g++, libstdc++ & I/O status


  I've spent some time to the g++ problem about I/Os.  I worked in many
directions with no success at all :-((

I'm using the simple program below to test g++ with standard io streams:

#include <iostream.h>
  cout << "hello" << endl;

First I tried to build this program with our current libg++ & g++  It compile clean but fails to run (it seg faults before displaying

Then I built libg++  No problem at compile time, but same behavior as
above at runtime :-(

I then tried to fix libg++  During compilation, it makes a bad choice
about providing or not its own libio implementation.  It should provide it
because g++ does not use thunks for formatting its virtual tables but
the libc6's libio implementation is based on these thunks.
After I fixed the configuration step to force libg++ to provide its own libio,
my test program ran a bit forward:  it displayed the message but segfaulted at
exit time :-(  It seems there is a pb in _IO_flush_all_buffered (or something
like that).

After these non productive trials, I chose to switch to egcc.

I then tried egcc 1.0 & its libstdc++ 2.8 from the experimental archive, with
no more success.

I further worked on egcc 1.0.2 (the latest release available including some
fixes for glibc 2.1).
Now my program hangs without displaying the message at all.
Under the debugger, it seems it loops indefinitely between _IO_file_seekoff and
__register_frame_info in the _init function (called before main).

I'm sorry, I have no more idea to provide a working g++ :-((
In effect I don't know if the bug is in g++, libstdc++ or libc6 :-(

Any new idea?

 Eric Delaunay                 | "La guerre justifie l'existence des militaires.
 delaunay@lix.polytechnique.fr | En les supprimant." Henri Jeanson (1900-1970)

To UNSUBSCRIBE, email to debian-sparc-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Reply to: