On Tue, 10.03.15 17:34, Alban Crequy (alban.cre...@gmail.com) wrote:
>
> - r = renameat2(AT_FDCWD, i->temp_path, AT_FDCWD,
> i->final_path, RENAME_NOREPLACE);
> + r = rename_noreplace(AT_FDCWD, i->temp_path, AT_FDCWD,
> i->final_path);
> if (r < 0) {
> r = log_error_errno(errno, "Failed to move RAW file
> into place: %m");
> goto finish;
If rename_noreplace() would return "-errno" like all the other calls
we define, then this would become:
if (r < 0) {
log_error_errno(r, "Failed to move RAW file into place: %m");
goto finish;
}
Which is both shorter and more inline with the rest of our code...
> - if (renameat2(AT_FDCWD, t, AT_FDCWD, to, replace ? 0 :
> RENAME_NOREPLACE) < 0) {
> + if (replace) {
> + r = renameat(AT_FDCWD, t, AT_FDCWD, to);
> + } else {
> + r = rename_noreplace(AT_FDCWD, t, AT_FDCWD, to);
> + }
> + if (r < 0) {
Please, no {} for single-line if blocks. See CODING_STYLE.
> + ret = unlinkat(olddirfd, oldpath, 0);
> + if (ret < 0)
> + unlinkat(newdirfd, newpath, 0);
Recently we started prefixing calls like this where we knowingly
ignore the return value with casts to (void). This tells code checkers
like Coverity that we *knowingly* ignore the error condition
here. Hence:
(void) unlinkat(newdirfd, newpath, 0);
instead of:
unlinkat(newdirfd, newpath, 0);
Lennart
--
Lennart Poettering, Red Hat
_______________________________________________
systemd-devel mailing list
systemd-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/systemd-devel
