[Crackerjack-devel] question about testcases/write

[Samuel Q. <Sam...@sy...>]

Hello,

I have crackerjack-3.3.  When I run the write test almost every test fails with "Bad file descriptor".  This is because most of the files are opened readonly – using FLAGS2 in lib/kor/test_common.c.

See for example symbolic_exist_file_fd() in test_common.c:

#define         FLAGS   O_RDWR
#define         FLAGS2  O_RDONLY

int symbolic_exist_file_fd()
{
        int             fd;
        char    tmpfile[MAXFILENAME];

        memset(&tmpfile, 0, MAXFILENAME);
        strcpy(tmpfile, symbolic_exist_file());

        fd = open(tmpfile, FLAGS2);

        errno=0;
        return  fd;
}

It seems to me like there should be a version of symbolic_exist_file_fd() that uses FLAGS instead of FLAGS2 for the write case.

Why does cs_write.c call symbolic_exist_file_fd()?  Look at this declaration in testcases/write/cs_write.h:

cs_write.h
struct data_and_gen fd_data[] = {
#include "fd_name_read.h"
};

include/kor/fd_name_read.h
{ 1, "EXIST_FILE_FD",  exist_file_fd },
{ 2, "EXIST_DIR_FD",  exist_dir_fd },
{ 3, "SYMBOLIC_EXIST_FILE_FD", symbolic_exist_file_fd },
{ 4, "SYMBOLIC_EXIST_DIR_FD", symbolic_exist_dir_fd },
{ 5, "PIPE_FD", pipe_fd },
//{ 6, "FIFO_FD", fifo_fd },    // be holding status
{ 7, "MAX_FD", max_fd },
{ 8, "INVALID_FD", invalid_fd },
{ 9, "CLOSE_FD", close_fd },
{ FINAL_DATA, NULL, NULL }


There is a similar issue with pipe_fd(), it returns the read end of the pipe.  Shouldn't there be a version that returns the write end?

-Sam