Some code fix suggestions in mountall.c

Binary package hint: mountall

Until I discover I'm affected by the bug #430542, I looked mountall's source code during several hours in order to learn how to get my system bootable again. I got my computer working again by changing my "/etc/fstab".

But... I think there are 2 things to be improved in mountall-0.1.6.

-------------------------------

The function "cleanup" (src/mountall.c:907) walks through the list of mount entries and removes the ones which have unknown file system type. The function "parse_filesystems" (src/mountall.c:848) discovers what file system types the kernel supports.

But there is a problem here: some file system types are implemented by kernel modules. These modules may not be loaded into kernel yet, so those file system types won't appear in /proc/filesystem. But, unfortunately, the function "parse_filesystems" don't try to check if unknown file system types could be implemented by not-loaded modules. As a result, "cleanup" drops mount entries which should be mounted.

This is my first suggestion: the function "cleanup" should try to load file system modules before discarding invalid mount entries.

-------------------------------

First, a code example:

// File "test.c" begin
#include <assert.h>
int main (void) {
    int test = 1;
    assert (test--);
    return (test);
}
// File "test.c" end

Build the code with these command lines:
$ gcc -o test_a test.c
$ gcc -DNDEBUG -o test_b test.c

Run these executables:
$ ./test_a ; echo $?
$ ./test_b ; echo $?

You will see the decrement is not executed when NDEBUG is defined (test_b).

Such problem may happen in several lines within mountall.c:
1165: nih_assert (i++ < num_paths);
1675: nih_assert (i++ < num_paths);
1717: nih_assert (i++ < num_paths);
2325: nih_assert (udev = udev_new ());
2326: nih_assert (udev_monitor = udev_monitor_new_from_netlink (udev, "udev"));
2327: nih_assert (udev_monitor_filter_add_match_subsystem_devtype (udev_monitor, "block", NULL) == 0);
2328: nih_assert (udev_monitor_enable_receiving (udev_monitor) == 0);

Perhaps, such errors don't happen now, because "nih_assert" is always compiled. But, in the future, some developer may implement a macro that affects nih_assert() just like NDEBUG affects assert()... If this day comes, the code will break...

This is my second suggestion: change the assertion lines. By example:

1165: nih_assert (i++ < num_paths);

Change to:

1165: nih_assert (i < num_paths); i++;