@abh3 commented on this pull request.
> @@ -2252,10 +2278,29 @@ int XrdOfs::rename(const char *old_name, // In
evsObject->Notify(XrdOfsEvs::Mv, evInfo);
}
+// If we cannot overwrite, we must open-exclusive first. This will test whether
+// we will destroy data in the rename (without actually destroying data).
+//
+ std::unique_ptr<XrdSfsFile> tmp_fp = nullptr;
+ if (cannot_overwrite)
+ {tmp_fp.reset(newFile(einfo));
+ if (!tmp_fp)
+ {return fsError(einfo, ENOMEM);}
+ if (SFS_OK != tmp_fp->open(new_name, SFS_O_CREAT, 0700, client, infoN))
Well, all of the side-effects of surreptitiously creating a file during rename certainly makes it unappealing. I'd really prefer an existence check. That way we get a reasonably predictable outcome regardless of what happens to the server. Frankly, the race condition hasn't been a problem in 20 years (yeah, famous last words).
--
Reply to this email directly or view it on GitHub:
https://github.com/xrootd/xrootd/pull/1697#discussion_r868680432
You are receiving this because you are subscribed to this thread.
Message ID: <[log in to unmask]>
########################################################################
Use REPLY-ALL to reply to list
To unsubscribe from the XROOTD-DEV list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-DEV&A=1
