LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Programming (https://www.linuxquestions.org/questions/programming-9/)
-   -   libarchive: dump memory archive to disk file (https://www.linuxquestions.org/questions/programming-9/libarchive-dump-memory-archive-to-disk-file-4175560398/)

atelszewski 12-02-2015 08:45 AM
libarchive: dump memory archive to disk file
 
Hi,

Libarchive version: 3.1.2

I'm creating the archive in memory with:
Code:
archive_write_open_memory(a, abuf, ABUF_SIZE, &asize);
Then I write the buffer out to disk like that:
Code:
write(fd, abuf, asize);
If I now try to unpack the archive in terminal, I get:
Code:
tar -xf output2.tar.gz

gzip: stdin: unexpected end of file
tar: Unexpected EOF in archive
tar: Unexpected EOF in archive
tar: Error is not recoverable: exiting now
On the other side, if I write the archive out to disk like that:
Code:
write(fd, abuf, ABUF_SIZE);
then it works as expected.

Can you please explain me what am I missing?

--
Best regards,
Andrzej Telszewski

atelszewski 12-02-2015 05:51 PM
Hi,

I did some code refactoring and it magically started to work correctly.
I don't know what happened and if it is going to be like that throughout the whole development cycle;)

Could somebody please confirm that the code I posted in the first post, should actually work and behave as expected (that is that with this code I should get correct on-disk archive)?

--
Best regards,
Andrzej Telszewski

norobro 12-02-2015 07:04 PM
You didn't show enough code to determine what might be wrong, so I modified the "man archive_write" example code to use "archive_write_open_memory" and it worked fine:
Code:
#ifdef __linux__
#define _FILE_OFFSET_BITS 64
#endif
#include <sys/stat.h>
#include <archive.h>
#include <archive_entry.h>
#include <fcntl.h>
#include <stdlib.h>
#include <unistd.h>

void write_archive(const char *outname, const char **filename) {
    struct archive *a;
    struct archive_entry *entry;
    struct stat st;
    char buff[8192];
    int len;
    int fd;
    size_t buff_size = 100000;
    size_t used;
    char *buffer;

    a = archive_write_new();
    archive_write_add_filter_gzip(a);
    archive_write_set_format_ustar(a);
    buffer = (char *)malloc(buff_size);
    if (buffer==NULL) {
        printf("Error allocating memory!");
        exit (1);
    }
    archive_write_open_memory(a, buffer, buff_size, &used);
    while (*filename) {
        stat(*filename, &st);
        entry = archive_entry_new();
        archive_entry_copy_stat(entry, &st);
        archive_entry_set_pathname(entry, *filename);
        archive_write_header(a, entry);
        if ((fd = open(*filename, O_RDONLY)) != -1) {
            len = read(fd, buff, sizeof(buff));
            while ( len > 0 ) {
              archive_write_data(a, buff, len);
                len = read(fd, buff, sizeof(buff));
            }
            close(fd);
        }
        archive_entry_free(entry);
        filename++;
    }
    archive_write_free(a);
    fd = open(outname, O_WRONLY | O_CREAT, 0644);
    write(fd, buffer, used);
    close(fd);
    free(buffer);
}

int main(int argc, const char **argv) {
    argv++;
    const char *outname = *argv;
    argv++;
    write_archive(outname, argv);
    return 0;
}
HTH

atelszewski 12-03-2015 02:07 AM
Hi,

@norobro thank you for confirmation. I based my code on the same example as you, plus some changes based on A Basic Write Example.

Is this part of any importance:
Code:
#ifdef __linux__
#define _FILE_OFFSET_BITS 64
#endif
?

Below comes my code, so if you have some comments, let me know. It's a C++ class, size and data are class members, for archive size and data buffer respectively.

Code:
/// Maximal archive size.
#define DATA_SIZE (1024U * 1024U)
/// Buffer size for file reading.
#define RBUF_SIZE (16U * 1024U)

/******************************************************************************/
int archive_t::create(filelist_t const &p_flst)
/******************************************************************************/
{
  struct archive *a = NULL;
  int res = 0;
  int m_open = ARCHIVE_FATAL;
  archive_entry *e = NULL;
  int rfd = -1;
  char rbuf[RBUF_SIZE];

  if (prgnam.length() == 0U)
  {
    emsg() << "archive: PRGNAM must be set." << endl;
    return -1;
  }

  if (p_flst.size() == 0U)
  {
    emsg() << "archive: file list is empty." << endl;
    return -1;
  }

  if (size != 0U || data != NULL)
  {
    emsg() << "archive: previous archive not freed." << endl;
    return -1;
  }

  a = archive_write_new();
  if (a == NULL)
  {
    emsg() << "archive: not enough memory." << endl;
    goto failure;
  }

  res = archive_write_set_format_pax_restricted(a);
  if (res != ARCHIVE_OK)
  {
    emsg_archive(a);
    goto failure;
  }

  res = archive_write_set_format_ustar(a);
  if (res != ARCHIVE_OK)
  {
    emsg_archive(a);
    goto failure;
  }

  data = new(std::nothrow) char[DATA_SIZE];
  if (data == NULL)
  {
    emsg() << "archive: not enough memory." << endl;
    goto failure;
  }

  m_open = archive_write_open_memory(a, data, DATA_SIZE, &size);
  if (m_open != ARCHIVE_OK)
  {
    emsg_archive(a);
    goto failure;
  }

  e = archive_entry_new();
  if (e == NULL)
  {
    emsg() << "archive: not enough memory." << endl;
    goto failure;
  }

  for (auto item = p_flst.begin(); item != p_flst.end(); ++item)
  {
    string apath;
    string rpath;
    ssize_t rsize;

    apath = prgnam + "/" + item->name;
    rpath =                item->name;

    archive_entry_set_pathname(e, apath.c_str());
    archive_entry_copy_stat(e, &item->stat);

    res = archive_write_header(a, e);
    if (res != ARCHIVE_OK)
    {
      emsg_archive(a);
      goto failure;
    }

    rfd = open(rpath.c_str(), O_RDONLY);
    if (rfd == -1)
    {
      emsg() << "archive: " << strerror(errno) << endl;
      goto failure;
    }

    rsize = read(rfd, rbuf, RBUF_SIZE);
    while (rsize > 0)
    {
      ssize_t wrote = 0;
      ssize_t wsize = rsize;

      while (wrote < rsize && wsize > 0)
      {
        wsize = archive_write_data(a, &rbuf[wrote], rsize - wrote);
        wrote += wsize;
      }

      if (wsize == 0)
      {
        emsg() << "archive: 0 size write." << endl;
        goto failure;
      }

      if (wsize == -1)
      {
        emsg_archive(a);
        goto failure;
      }

      rsize = read(rfd, rbuf, RBUF_SIZE);
    }

    if (rsize == -1)
    {
      emsg() << "archive: " << strerror(errno) << endl;
      goto failure;
    }

    close(rfd);
    rfd = -1;
    archive_entry_clear(e);
  }

  archive_entry_free(e);
  e = NULL;

  res = archive_write_close(a);
  m_open = ARCHIVE_FATAL;
  if (res != ARCHIVE_OK)
  {
    emsg_archive(a);
    goto failure;
  }

  res = archive_write_free(a);
  a = NULL;
  if (res != ARCHIVE_OK)
  {
    emsg_archive(a);
    goto failure;
  }

  return 0;

failure:
  if (rfd != -1) close(rfd);
  if (e) archive_entry_free(e);
  if (m_open == ARCHIVE_OK) archive_write_close(a);
  if (data != NULL) delete[] data;
  if (a) archive_write_free(a);

  data = NULL;
  size = 0U;

  return -1;
}
Code:
/******************************************************************************/
int archive_t::dump(string const &p_name)
/******************************************************************************/
{
  int fd;
  ssize_t wsize = size;
  size_t wrote = 0U;

  if (p_name.length() == 0U)
  {
    emsg() << "archive: name must be longer than 0." << endl;
    return -1;
  }

  if (data == NULL || size == 0U)
  {
    emsg() << "archive: invalid data." << endl;
    return -1;
  }

  fd = open(p_name.c_str(), O_WRONLY | O_CREAT | O_TRUNC,
            S_IRUSR | S_IWUSR | S_IRWXG | S_IROTH);
  if (fd == -1)
  {
    emsg() << "archive: " << strerror(errno) << endl;
    return -1;
  }

  while (wrote < size && wsize > 0)
  {
    wsize = write(fd, &data[wrote], size - wrote);
    wrote += wsize;
  }

  if (wsize == 0)
  {
    emsg() << "archive: 0 size write" << endl;
    close(fd);
    return -1;
  }

  if (wsize == -1)
  {
    emsg() << "archive: " << strerror(errno) << endl;
    close(fd);
    return -1;
  }

  close(fd);

  return 0;
}
--
Best regards,
Andrzej Telszewski

norobro 12-03-2015 09:25 PM
Quote:
Originally Posted by atelszewski
Is this part of any importance:
Depends on the systems you target. From "man feature_test_macros":
Quote:
_FILE_OFFSET_BITS
Defining this macro with the value 64 automatically converts references to 32-bit functions and
data types related to file I/O and file system operations into references to their 64-bit coun‐
terparts. This is useful for performing I/O on large files (> 2 Gigabytes) on 32-bit systems.
(Defining this macro permits correctly written programs to use large files with only a recompi‐
lation being required.) 64-bit systems naturally permit file sizes greater than 2 Gigabytes,
and on those systems this macro has no effect.
I cobbled together a header for your class and a main and your code compiled and ran fine.

I'm a little "goto" averse, so I would probably use try, throw, catch to handle errors.

atelszewski 12-04-2015 06:02 PM
Hi,

First of all thank you for taking the time to test the class in operation, I really appreciate that. Since then I slightly cleaned up the code and removed what seemed to be excess of code.

When it comes to _FILE_OFFSET_BITS I see I won't need it. My program is going to be run only on Linux and archives are going to be very small in size.

Quote:
I'm a little "goto" averse, so I would probably use try, throw, catch to handle errors.
There are two things. I haven't programmed in C++ in something like 5 years, but I'm programming a lot in C, so the "C" way is safer at this stage. Second thing is that, I also used to be extremely-anti "goto", but since then I loosen a bit the rules. Code written properly using "goto" is much better that incorrectly written code using "throw, try & catch". That being said, it doesn't matter much which "technology" you use as long as you design and code correctly, actually I think that both methods present the same risks.

I'm aware that throw & company would be more of C++ way of doing the things but, as I said, my knowledge is limited, so I prefer to use the known subset of C++.

Thanks for the help!

--
Best regards,
Andrzej Telszewski

atelszewski 12-08-2015 09:33 AM
Hi,

I'm marking as solved. It seems I had an error somewhere, other than that the solution is correct.

--
Best regards,
Andrzej Telszewski


All times are GMT -5. The time now is 03:10 AM.