pdaXrom 1.1.0 undefined symbol: __udivsi3 ???

[damiandixon]

Had this problem while porting code to embeded linux (ARM based but not zaurus).

I don't have the details at home, but you need to setup the LD_PRELOAD environment variable to load a number of shared libraries such as:


export LD_PRELOAD=/usr/lib/libgcc.so


You may need to also preload the shared math library.

If nm is available then you should be able to find which shared library you need to preload.


--
Still trying to convince the Wife to let me have a C860!

[lardman]

Does the above fix work?

Do any of the programs which people were having troubles with now work with RC3?

Cheers,


Si

[pmf]

Does the above fix work?

Do any of the programs which people were having troubles with now work with RC3?

At least Vim-6.2 (which gave the undefined symbol: __divsi3 error before) works, but you have to use

Code: [Select]

export LD_PRELOAD=/lib/libgcc_s.so.

Thanks to damiandixon for providing the solution; can't have a Zaurus without Vim  

[stbrock]

The fix did not work for me trying to compile pari-gp on a C860 with RC3. I still get:

      hidden symbol '__udivisi3' . . . .  is referenced by DSO

after about 10 minutes of compiling. I found that /lib/libgcc_s.so was not executable while the other similar files in /lib were, but changing it to be executable made no difference. I'm new at this, so I could easily be doing something else wrong. Must other compile options be set? I assumed that the options and instructions for cross compiling on the pdaXrom site for the earlier SDK were not applicable to the new native SDK. Are some of them needed in any event?

RC3 seems terrific in general, but if sources generally don't compile without a lot of effort, I'll probably have to head back to pocketworkstation and apt-get, which got pari-gp and several other programs running without any difficulty.  But if I can get a few standard options in place and then compile most programs, pdaXrom looks much better.

[lardman]

Going OT:

Is pari-gp not available in binary form already (http://wotan.algebra.math.uni-siegen.de/~countnumber/zaurus/ and http://modular.fas.harvard.edu/ipaq/).

That said, it may be linked against a slightly later version of libc (it's said to run on OZ3.2 which is libc 2.2 based, but pdaXrom seems to use an older subversion so ymmv).

Edit: I suppose you may also want to re-compile to enable X11 plotting, in that case there are some compilation tips in the first link.


Si

[stbrock]

Thanks for the leads, lardman. I had tried another binary, the .deb package, and compiling unsuccessfully. The binary at the first link failed for me in RC3 complaining about libqpe.so.1. (I suspect that I'm not the only newbie confused about the status of Qt under pdaXrom and that QtConfig tool.)

But the second link worked, so at least I have the gp calculator working on the Z, if not the graphing. It's a kick to see it spit out the exact several hundred digit value of 3578! almost instantly. As long as you keep away from floating point calculations, the Z is actually *fast*.  

I'll keep working on learning the Z compiling rituals with the hints in the first link as I get the time. I should probably hold off on posting too many questions about compiling under pdaXrom till people have a chance to work with it under the final 1.1 release.