Basic OS gives errors

[gsingh2011]

I am using the exact code for the loader.s, kernel.c, and linker.ld that you can find here:http://wiki.osdev.org/Bare_Bones. When I run the commands, I get this:


I am using Cygwin on Windows XP. The commands ran in the bash file compile.sh are:

Code: Select all

cd nasm-2.06rc1
./nasm -f elf -o ../loader.o ../loader.s
cd ..
gcc-4 -o kernel.o -c kernel.c -Wall -Wextra -nostdlib
ld -T linker.ld -o kernel.bin loader.o kernel.o

Anyone know why I'm getting those errors?

[xDDunce]

hiya,

first thing i see is there should be an underscore infront of kmain making it _kmain, so you need to edit your loader.s to make it so.

and as for the warnings, they're just warnings. ignore them. unless you are using a strip program, they don't mean much to be honest.

if this doesnt work then i guess i missed something.

cheers,
James

[Combuster]

For undefined reference to kmain:
* use a GCC Cross-Compiler
* use -fno-leading-underscores
* add underscores for each c function you call from assembly

they're just warnings. ignore them.

Warnings indicate bugs, potential bugs, bad coding practice or design, so you should fix them, not ignore them. Let alone tell others to do so

[JJeronimo]

they're just warnings. ignore them.

Warnings indicate bugs, potential bugs, bad coding practice or design, so you should fix them, not ignore them. Let alone tell others to do so

You forgot to mention that a console window full of warnings is ugly.

JJ

[AJ]

You forgot to mention that a console window full of warnings is ugly.

Indeed - no news is good news . I suggest using the -W flags from Solar's Makefile tutorial:

Code: Select all

-Wall -Wextra -pedantic -Wshadow -Wpointer-arith -Wcast-align \
              -Wwrite-strings -Wmissing-prototypes -Wmissing-declarations \
              -Wredundant-decls -Wnested-externs -Winline -Wno-long-long \
              -Wconversion -Wstrict-prototypes

Cheers,
Adam

PS: If adding them to a big project, add them one at a time

[xDDunce]

they're just warnings. ignore them.

Warnings indicate bugs, potential bugs, bad coding practice or design, so you should fix them, not ignore them. Let alone tell others to do so

yes... i actually meant just the ones stated. i see no way of correcting them, and i doubt anyone else does without adding useless, functionless, time-wasting, space-filling code.

and before you say it... yes, i need to think more before posting.

cheers,
James.

[JJeronimo]

yes... i actually meant just the ones stated. i see no way of correcting them, and i doubt anyone else does without adding useless, functionless, time-wasting, space-filling code.

In this case, you are right.
Sincerely speaking, I didn't know GCC warned against unused parameters. I thought it would only warn against unused (non-parameter) local variables.

JJ

[AJ]

i see no way of correcting them, and i doubt anyone else does without adding useless, functionless, time-wasting, space-filling code.

Code: Select all

int main(int /* argc */, char** /* argv */) 
{
   return(0);
}

Cheers,
Adam

[xDDunce]

i see no way of correcting them, and i doubt anyone else does without adding useless, functionless, time-wasting, space-filling code.

Code: Select all

int main(int /* argc */, char** /* argv */) 
{
   return(0);
}

Cheers,
Adam

which, again, is bad practice. the kernel should never return. especially if you're using C++ because of the destructors list.

also the fact that the parameters defined there are commented out so they wont be compiled, and that therefore means the variables don't actually exist. resulting in an incomplete multiboot header.

cheers,
James.

[JJeronimo]

which, again, is bad practice. the kernel should never return.

Are you sure? I think this is just an implementation detail.
The main() can be just the "bootup driver[1]", which falls out of scope once the multitasking stuff is on.
And it's not the kernel that is returning. It's just the main() function, which happens to be in the kernel binary.

JJ

[1] A driver is someone/thing that drives. Not necessary in the "device driver" sense.

[xDDunce]

even so, isn't a driver just another program but with higher priviliges than just plain old user mode? just like the kernel. so returning from any main() function (be it main(), kmain() or anything similar) should not be done as it will cause a termination of the program.

cheers,
James.

[Combuster]

In kernel space, there is no C library, no runtime. What main does when it returns is not defined until YOU program it.

[xDDunce]

In kernel space, there is no C library, no runtime. What main does when it returns is not defined until YOU program it.

and by standard (standard being what the general consensus [can't spell ] does), that will be your destructors list (if using C++) which kills all global classes. which, if done unintentionaly, will cause triple faults left, right and center. so technicly, you should return when shutting down (if at all), to ensure all important data is written to disk and such. followed by the actual shutdown sequence. it's the fastest, simplest and most reliable way. and from what i see, most (if not all) tutorials do it, along with many of the open source projects i have taken a look at recently.

[abachler]

make sure you are using the cross compiler version, as the gcc you download is not compiled as a cross compiler, you have to specificalyl recompie it to be a true cross compiler, otherwise it will mangle the function names (as it should).

[quok]

i see no way of correcting them, and i doubt anyone else does without adding useless, functionless, time-wasting, space-filling code.

Code: Select all

int main(int /* argc */, char** /* argv */) 
{
   return(0);
}

Cheers,
Adam

Even better, but GCC specific:

Code: Select all

int main( int argc __attribute__((unused)), char *argv __attribute__((unused)) )
{
    return 0;
}