Fix GString API migration regressions: lack of NUL after strncpy and use after free

[peripherium]

strncpy() was called with text->len, which does not include the terminating NUL byte. As a result, when a shorter filename was copied after a longer one, the global buffer retained trailing characters from the previous value, producing an invalid path.

The fix copies text->len + 1 bytes to include the guaranteed NUL terminator provided by GString.

[mc-worker]

Fixes 260126a.

[mc-worker]

/rebase

[zyv]

I wish we had a test for this function, especially when a bug is getting fixed.

@peripherium, would you be up for the task? I have added a test under similar circumstances; you can check this as an example: 7488fc5 .

[KuzinAndrey]

I build mc from last commit 9f0be66 and now can't open any tgz archive (after open one). Always get an error:

But this fix is help to return to work.

[peripherium]

Hi,

while adding the requested unit test for exec_make_shell_string(), I ran into a small question regarding the intended testing pattern.

The function was declared static, so I exposed it to the tests using MC_TESTABLE. However, it also relies on the global variable buffer, which is likewise declared static in the same C file.

To make the test build cleanly with and without --enable-tests, I added the following in the forward declarations section in ext.c:

// To prevent compiler warnings from being generated with --enable-tests
MC_TESTABLE
GString *exec_make_shell_string (const char *lc_data, const vfs_path_t *filename_vpath);

// To prevent compiler warnings from being generated without --enable-tests
#if defined(HAVE_TESTS)
extern char buffer[BUF_1K];
#endif

This works for the test, but I wanted to ask if this is the preferred approach in MC, or if there is an established pattern for handling static globals used by testable functions.

If there is a better way to structure this, I’d be happy to adjust the patch.

Thanks.

[mc-worker]

@zyv let's merge this PR. Currently, in master, we have broken entering to archives.

[mc-worker]

I found another regression in 260126a. The second commit in this PR fixes it.

@zyv let's merge this PR to fix the broken extfs in master.

[zyv]

@zyv let's merge this PR to fix the broken extfs in master.

Sorry, I'll try to have a look today.

[zyv]

/rebase

[zyv]

Okay, guys, I'm sorry that it has taken me so long.

I have fixed the build on macOS due to duplicate symbols. Otherwise, I find the code good. And I appreciate the test very much. Thank you @peripherium for an excellent contribution!

I will not be adding this to the NEWS, since it's a regression in master after an internal migration.