In our project, we are building an ELF file and a partially linked file (PLF) which is converted to a proprietary format and loaded into memory after the ELF is loaded. We use Codewarrior to run and debug, which has been working just fine (the C++ source code is always available to step through when debugging).
I've recently made a change where some code and data are compiled into a different section in the PLF file (.init, which was previously empty). Now, when debugging, a majority of the files are available only in assembler. When I re-build, no longer using .init, we can step through C++ source code again.
Does anyone know why this would be the case?
why this would be the case
One reason could be that codewarrior is not expecting to find code in .init section.
You are unlikely to get a good answer here. Try codewarrior support forums.
I got this working by switching the order of the sections using the linker command file (.lcf) so that the .init section comes second after .text. I guess as Employed Russian suggests, CodeWarrior is surprised by having code in .init and craps out. Changing the order of the sections seems to have no ill effects and now debugging works as expected again.
Related
I'm trying to create some debug scripts with compiled programs, for this I'm trying to create something where I prepare my variables in some code I generate and then jump into another program.
Is there a way to do that ? For example by having some C code and then jumping to a label or place in the executable. For now I'm focusing on ELF programs, but if something exists on Windows I'm also interested !
Thanks !
I've tried to bring back the ELF file into a .s for GCC and recompile, however this doesn't seem to work well for all ELF files (e.g non-PIE binaries). And I've looked to see if there were tools that would create a .s but they are either buggy, incomplete or both.
Background:
I am generating a JIT code (which generates x86-64 code). After the end of JIT process, I have a .text section, a .data section and a .eh_frame section generated (.eh_frame is used for stack unwinding). I am able to execute this JIT code successfully. But the issue is GDB. I want to be able to debug this JIT code using GDB (specifically the 'backtrace' command of GDB should work).
Problem:
I need to tell GDB about this loaded JIT code (in particular I need to tell GDB about .eh_frame so it can use that frame for stack-unwinding). I see that GDB has a JIT interface: https://sourceware.org/gdb/current/onlinedocs/gdb/JIT-Interface.html
Possible Solutions:
There are two options here:
Hand over a ELF file to GDB
Write a Customer Jit-Reader plugin to handle debugging of custom object file.
Right now I have a custom object file (just bunch of three independent sections loaded into memory). I don't want to write my own Jit Reader plugin.
Blocking Issue:
Does anyone knows existing code that will help me package these three independent sections into a simple ELF file (which I can then register with GDB by calling __jit_debug_register_code())? I am guessing all I need to do is write some header (conforming to ELF specifications) which have names and pointers to the section. Is there existing open source code for this or if not, can someone point me towards how to do this packaging myself?
I need bare minimum ELF file so that GDB is happy (I don't need to Load the ELF file as .text and .data section are already loaded)
libelf could be of help for constructing an ELF object. There are open-source implementations available at:
elftoolchain (BSD licensed)
elfutils (GPL).
I'm trying to do a hello world for a boot loader in assembly.
I'm following this tutorial:
http://www.osdever.net/tutorials/view/hello-world-boot-loader
I searched and it seems people are saying gcc doesn't work for compiling assembly. So I searched and found flat assembler. When I try to compile the example, it gives me an error at the first line [BITS 16]. Basically it states 'Illegal Instruction'.
What type of assembler does this code require?
I don't know if some tool in GCC can compile assembler (and if it does, whether it can compile to 16 bit code), but the site you refer to recommends NASM. Did you try it with NASM?
I simply commented the line out and it worked in FASM. It seems like FASM defaults to 16 bit automatically. After it compiled, it generated a BIN file by the same name. I renamed it to an IMG extention and then assigned it as a Floppy Disk image using VirtualBox to test it. Worked great and booted.
As long as you write this BIN/IMG file to the drive on the first sector it seems to work okay. I used the tutorials on the above website also.
Try removing square brackets around BITS 16 in case you didn't.
I recently compiled GotoBLAS2 (MacOSX 10.6) and linked it to my code, leading to all kind of wrong results. I ran everything through valgrind noticing some illegal reads from the GotoBLAS. When looking at it more carefully I found that the GotoBLAS is compiled with the -m128bit-long-double alignment option. As soon as I did compile my code with this flag as well (although I don't use any long doubles at all) everything works, giving correct results without any valgrind obscurities.
Now my question ist:
Do I have to compile all other library dependencies using the same alignment flag?
I'm working on a project, and on a machine without Xcode, I'm getting a crash. (of course it works on my machine B-/) I have a crash log, with a PC offset for the crash. I'd like to be able to see where that actually is in the code. I know that Code Warrior can disassemble the code (presumably, only debugable code) and show it interspersed with the C code, then I just have to look for that address, and I'm done.
Is there an easy way to do this in Xcode?
thanks.
There are two things in Xcode you may want to look at. The first would be to select your source code file and choose Build->Show Assembly Code. What this won't give you though is offsets.
The second assembly capability is in the debugger. Choose Run->Debugger Display->Source and Disassembly, and the debugger will show you both source and assembly code side-by-side. However, the two are not interspersed.
If neither of these Xcode facilities give you what you need, your only recourse may be the otool command line tool.
I've never found a way to generate or view source and assembly interspersed.