Hi,
The saga continues.
Today, I've updated my Fedora 9 "Sulphur" installation to Fedora Rawhide (development version) in order to close the gap with Debian Lenny.
My program is now also crashing with Fedora Rawhide due to the incorrect behaviour of atof!
So many packages have been updated that I cannot determine which one is the culprit. I've nevertheless tried to downgrade to the glibc, libstc++ and gcc packages provided by Fedora 9 "Sulphur", but it didn't help.
To summarize, the problem isn't limited to Debian and I don't know if it can be reproduced on a non-ia64 system.
Best regards,
Émeric
2008/8/10 Émeric Maschino
<emeric.maschino@gmail.com>
Hi,
Today, I ran on the Fedora installation the binary produced with the Debian installation. Guess what? Yes, it worked like a charm!
Now the strange thing. I ran on the Debian installation the binary produced with the Fedora installation. Segfaults due to the atof returning 0 rather than 0.189.
So, the problem isn't at compile time, but at runtime, and only on the Debian installation. I'm still wondering what's going wrong, since a simple C++ program with a single call to atof runs flawlessly on my Debian installation.
Best regards,
Émeric
2008/8/9 Émeric Maschino
<emeric.maschino@gmail.com>
Hi,
Today, I recompiled the whole project on my Fedora installation using the updated gcc 4.3.1-4 packages from Rawhide. No problem with gcc 4.3.1 on my Fedora installation.
By contrast, despite a complete recompile with gcc 4.3.1 on my Debian Lenny installation, atof and strtof still behave incorrectly!
Is there any significant patch applied by Debian on the gcc 4.3.1 upstream code (or any related package) that could explain this situation?
Best regards,
Émeric
2008/8/8 Émeric Maschino
<emeric.maschino@gmail.com>
Hi,
I have free time to play Fedora 9 "Sulphur" on my hp workstation zx6000.
The exact same C++ project (since the source code is on a separate partition) runs as expected on Fedora.
gcc version is 4.3.0 20080428.
My Debian Lenny installation comes with gcc
4.3.1.Has something changed between gcc 4.3.0 and 4.3.1 regarding the standard C/C++ header inclusion?
I'm really running out of ideas...
Best regards,
Émeric