fix(logstorage): replace unbounded strcat with snprintf in storage_dir_info

[SoundMatt]

Summary

dlt_logstorage_storage_dir_info() in dlt_offline_logstorage_behavior.c builds a file path from three pieces using three chained strcat() calls into a fixed-size stack buffer:

char tmpfile[DLT_OFFLINE_LOGSTORAGE_MAX_LOG_FILE_LEN + 1] = { '\0' };  // 121 bytes
if (dir != NULL) {
    strcat(tmpfile, dir);          // up to ~99 chars (from config)
    strcat(tmpfile, "/");          // + 1
}
strcat(tmpfile, files[i]->d_name); // up to NAME_MAX = 255 chars

dir is derived from a config-bounded field (max DLT_OFFLINE_LOGSTORAGE_MAX_FILE_NAME_LEN = 100 chars), so the directory component is bounded. However files[i]->d_name comes from scandir() and can be up to NAME_MAX (255 bytes) on Linux.

dir(~99) + "/" + d_name(~255) = ~355 bytes can easily overflow the 121-byte tmpfile buffer, causing a stack buffer overflow. The overflow is triggerable whenever the storage directory contains a file whose name exceeds the buffer capacity, which can happen naturally or via a crafted mount-point with long filenames.

Changes

Replace the three strcat() calls with a single snprintf(..., sizeof(tmpfile), ...), which hard-limits the write to the buffer size. Add a strncpy for the no-directory branch for consistency.

Testing

• Paths shorter than DLT_OFFLINE_LOGSTORAGE_MAX_LOG_FILE_LEN are copied unchanged.
• Paths longer than the buffer are safely truncated; no stack overflow.

[minminlittleshrimp]

Please kindly check and fix the api logic, not remove it completely.
Thank you

[minminlittleshrimp]

Hello @SoundMatt
please kindly explain why the official api is removed.
Kindly approach with fixing the official API instead of manually handly app id or ctx id.

[SoundMatt]

Thanks @minminlittleshrimp — I think the confusion was that this branch was accidentally stacked on top of the GET_LOG_INFO V2 commits, so your comment landed on the get_log_info_v2 apid/ctid handling at dlt_daemon_client.c:2154 rather than the logstorage change.

I've rebased this PR onto master so it now contains only the logstorage fix (the single storage_dir_info strcat→snprintf change). No official API is touched here.

The get_log_info_v2 apid/ctid parsing you flagged has been reworked to use the shared dlt_set_id_v2() helper (into local buffers, no raw wire-buffer aliasing) over in its own PR #861 — same approach as #864/#868. Could you re-review the two separately now that they're cleanly split?