My turn - GCC 3.3.1 - Internal compile error (Panicking)

Trying to build your own version of Vega Strike and having problems? Unix users, paste your config.log here (stderr output alone is not helpful).
Post Reply
klauss
Elite
Elite
Posts: 7243
Joined: Mon Apr 18, 2005 2:40 pm
Location: LS87, Buenos Aires, República Argentina

My turn - GCC 3.3.1 - Internal compile error (Panicking)

Post by klauss »

Ok... It's my turn to call for help.

After several hours trying to get ./configure running (and succeeding - yay):

Linux version 2.4.21-297-smp4G (root@i386.suse.de) (gcc version 3.3.1 (SuSE Linux)) #1 SMP Sat Jul 23 07:55:00 UTC 2005

When compiling hard_coded_scripts.cpp,
in boost129/boost/python/detail/referent_storage.hpp, line 72:

Internal compiler error: Segmentation

Besides changing GCC versions... any idea?
GCC 3.3.1 is pretty common, I don't think that should be happening.

Besides, I'm not sure I can change the compiler... that's running on the server at work. Using up CPU resources for compiling is one thing... not that they're short on CPU power... but messing with software is another.

PS: I'll post the exact output as soon as I get another chance at building. Configure runs smoothly - no prob at all.
Oíd mortales, el grito sagrado...
Call me "Menes, lord of Cats"
Wing Commander Universe
Wisq
ISO Party Member
ISO Party Member
Posts: 453
Joined: Sat Jul 30, 2005 10:21 am

Post by Wisq »

Try just running 'make' again. That's what I did. You might get more segmentation faults, but they should hopefully occur at different points, and eventually, you'll have a complete build.

You may also consider running "memtest86+" on your system. I had a similar problem (segfaults), and some devs were suggesting it was all about the version of the "boost" library they used. However, I also had a bad section of my RAM, and this caused other failures as well, so I still don't know if all my GCC errors were just the same memory glitch.
klauss
Elite
Elite
Posts: 7243
Joined: Mon Apr 18, 2005 2:40 pm
Location: LS87, Buenos Aires, República Argentina

Post by klauss »

I somehow doubt both.

First, I tried running make many times. The error appeared always at the same spot. Then I tried rewriting that section of code in some other equivalent way (in MSVC, at least, that usually gets around internal compiler errors).
Still, the error stayed at the same place, immune to all my tests.

That also makes me think it's not the memory: after all I did, the outcome should have changed. Anyway, to be really sure about the memory I should do the test: I'll try on Monday (when I power up the server). But that would be very odd as well: all the processes running in it (databases of all kinds, ciphered disk volumes, etc...) run perfectly well 24-7. (well, 24-5 actually).

But there's the other thing: I know many devs (hellcat himself, IIRC) use GCC 3.3 to build, without problems. I'm not sure it's 3.3.1, though. Perhaps there's another subversion.
Oíd mortales, el grito sagrado...
Call me "Menes, lord of Cats"
Wing Commander Universe
ace123
Lead Network Developer
Lead Network Developer
Posts: 2560
Joined: Sun Jan 12, 2003 9:13 am
Location: Palo Alto CA
Contact:

Post by ace123 »

You can try compiling the file with problems directly from the command line:

Take the last gcc ... ... ... command that was printed out (usually 3-4 lines or so) and copy it to the prompt (you might want to paste in a text editor to make sure that there are no new-lines in the middle.

Then, try removing optimization flags (-f... and -O2)

or change the -O2 to a -O1
klauss
Elite
Elite
Posts: 7243
Joined: Mon Apr 18, 2005 2:40 pm
Location: LS87, Buenos Aires, República Argentina

Post by klauss »

Ok. I did run the memory test, and it's all ok.
I also did try fiddling with all kind of optimization options, adding and removing things like -O1, -O2, -O3, -O0, -pipe, etc...

The -Ox things moved the error around, but didn't stop it from occurring. The others had no effect. I didn't try running it from the commandline, though. I did all that modifying the Makefile, so I'll have to try that later.

PS: Anyone feeling like replacing boost131 by boost132? I heared it fixes this specific error. What puzzles me, though, is that I've heared others being able to compile under gcc 3.3 *with* boost 131. So I'm feeling a tad jealous.
Oíd mortales, el grito sagrado...
Call me "Menes, lord of Cats"
Wing Commander Universe
Post Reply