iLiad Has anyone ported ltrace to the iLiad?

[LazyScot]

The title says it all

I know the code is easy to get, but it has a bunch of dependencies, including elf library support stuff that are not currently on the iLiad, so before I go through the process of trying to get them compiled I was wondering if anyone had already done it (to confirm it will work) and would be willing to make it available...

[Antartica]

Originally Posted by LazyScot

The title says it all

I only had compiled strace. I wasn't aware of the existance of ltrace (seems a nice complement to strace :-b).

And it only seems to depend on libelfg0, so it will be easy to port (see: http://packages.debian.org/lenny/ltrace )

(a while later)
I've undusted my iliad development and compiled it, with the only dependecy statically linked for convenience.

You can grab it here:

http://projects.mobileread.com/iliad...5_iliad.tar.gz


P.S.: I've also updated the dl-and-install.sh script of my toolchain, as it didn't work anymore. More info here: https://www.mobileread.com/forums/sho...924#post216924

P.S.2: What I did for this is (for future reference ;-) ), in the scratchbox environment:
$ ./dl-and-install.sh libelf.so.0
$ ./dl-and-install.sh usr/bin/ltrace
$ cd ltrace-0.5/
$ ./configure
$ make
$ rm ltrace
$ gcc sysdeps/sysdep.o ltrace.o options.o elf.o output.o read_config_file.o execute_program.o wait_for_something.o process_event.o display_args.o breakpoints.o proc.o demangle.o dict.o debug.o summary.o ../libelf-0.8.10.orig/lib/libelf.a -lsupc++ -liberty -o ltrace
$ tar -cvzf ltrace-0.5_iliad.tar.gz ltrace

P.S.3: I've not tested it on the iliad, so I don't know if it works...

[LazyScot]

Many thanks for this.

I've grabbed your binary, but it seems to have the same problem I'd run into last night when I made a quick try to build it (though that wasn't in the sandbox). On in the iLiad, it seems to get stuck in an infinite loop calling __libc_start_main. I had suspected that this was something to do with either the libelf not allowing it to patch up the call/return stack correctly once it has traced the call, or the cross compilation not going right -- ltrace doesn't really cross compile (I'd had to hack a few files). Given you've done it is the simulator, its not a cross compile problem -- I probably now need to start understanding the code .

But not, I think, tonight -- I'll add an update if I make progress (though if someone has it working, let me know -- it is possible I've managed to break something else on the iLiad....)

[Antartica]

Originally Posted by LazyScot

Many thanks for this.

You're welcome

Originally Posted by LazyScot

ltrace doesn't really cross compile (I'd had to hack a few files). Given you've done it is the simulator, its not a cross compile problem -- I probably now need to start understanding the code .

The part you're talking about is the one in the sysdeps directory. The README there explains that those are the architecture-specific functions that do the hard work.

While compiling it I feared that configure detected that I was using a 2.6 kernel and compled the worng parts, but after looking around, I didn't saw any hints for different code in 2.4 and in 2.6. BUT when I tried ltrace in scratchbox, I get a lot of lines such as:

qemu: Unsupported syscall: 26

So, may be it's kernel dependant. Anyway, I've compiled an older version (0.3.36, the one that came with debian sarge, as sarge used a 2.4.x kernel), if you can try if it works... ;-).

http://projects.mobileread.com/iliad...6_iliad.tar.gz

P.S.: ltrace 0.3.36 generates the same unsupported syscall as ltrace 0.5, so my hopes are not very high...