Thread: Help with recoding Tar command

1.) What I do is, read a file, stat() on the file, create a header for my tar file, write into the tar file. And after I do the same for the rest. Is this more or less correct?

Sure, doesn't sound too far off the mark . . .

2.) Im having trouble with executables. I open an executable with emacs, you see very strange things. If I copy and paste this into a new file, it does not want to run no matter the encoding I put. Also reading the executable, the "read" command, doesnt read beyond the first 10 characters no matter what.

Try opening your file in binary mode. You can do this by appending a "b" to whatever fopen() mode you are using.

3.) If you open with emacs the executable or a tar file, you see ^@^@^@^@^@^@ and stuff like that all over the place. This is what a tar command (for example) does to make sure everything is in blocks of 512. How do I imitate this? I cant just write ^@^@ as it doesnt work. (im guessing ^@ is some binary code).

^@ is a NULL, I believe. Character 0.

^A is 1, ^B is 2, ^Z is 26, escape is 27, and ^@ is 0, as far I can remember. (All in decimal.)

Would opening in binary mode, and reading it, work the same as usual opening and reading (just that it will preserve the binary characters?)

For the header: Im confused with the checksum field.
"The chksum field is a checksum of all the bytes in the header, assuming that the chksum field itself is all blanks."

How would I go about finding that out?

How can I fill a string with "zero" bytes?

To begin with, why don't you write a simple program that copies a file in binary mode? Yes, it should work to do that. You will, of course, have to also set the correct mode of the file, and so on when you create the new file.

By the way, I admire your courage. tar is not an entirely simple product, and I think it would take me quite some time to get such an application put together and working correctly. Not to mention that there are a great number of flags that make tar do things ever so slightly differently. Bet that as it may.

Setting some data structure to zero is easiest done with

Code:

memset(&something, 0, sizeof(something));

--
Mats

Ive written a program that copies a file into another file, it works both for executables and others. (I did not use binary mode by the way, I used the example given in the FAQ).

Right now Im in the process of trying to write a function that will create the header. Not too much sucess so far, but Im getting closer. The problem is writing the checksum (which I dont know where to get it), and filling the various strings with Zero bytes. (bzero is a good function also?)

EDIT: I still have problems with writing the checksum. No idea how to get it.

What have you done so far with regards of checksum?

bzero isn't as portable as memset, but yes, you can use bzero if you don't care about being portable. I would recommend memset, as I'm not aware of ANY OS that doesn't have a memset [it's probable one of the most common functions in <memory.h>, possibly after memcpy, so it's a quite common function]. I'm pretty sure that bzero, on systems that support it, is implemented like this:

Code:

void bzero(void *ptr, size_t size)
{
    memset(ptr, 0, size);
}

If you really want to avoid writing the third argument, why not make a macro like this:

Code:

#define Zero(x) memset(x, 0, sizeof(x))

--
Mats

According to wikipedia:

Originally Posted by Wikipedia tar file format

The checksum is calculated by taking the sum of the byte values of the header block with the eight checksum bytes taken to be ascii spaces (value 32). It is stored as a six digit octal number with leading zeroes followed by a nul and then a space.

So you shouldn't set the checksum to all zero, but something like:

Code:

   memset(checksum, 32, 6);

--
Mats

I would recommend memset, as I'm not aware of ANY OS that doesn't have a memset [it's probable one of the most common functions in <memory.h>, possibly after memcpy, so it's a quite common function].

Don't you mean <string.h>? . . . memory.h is an extension with the mem*() functions from string.h, as far as I can tell. I do know that string.h is standard, and I'm pretty sure that memory.h isn't.

Originally Posted by dwks

Don't you mean <string.h>? . . . memory.h is an extension with the mem*() functions from string.h, as far as I can tell. I do know that string.h is standard, and I'm pretty sure that memory.h isn't.

Yes, I mean string.h of course.

--
Mats

Hey again!

All is advancing well. I succeeded in creating the header, and tar files created with my program can be extracted with the tar command.
Im looking into being able to tar Directories in a recursive manner. Would I be able to adapt the Walker located on the FAQ for this question? Or would alot of it be useless for me?

Thanks guys!