[tuffnatty]

Proposed changes

The "full file cloning" copy path that is taken on MacOS and Oracle Solaris requires that the destination does not exist before the cloning attempt. Removing the destination before cloning has various undesirable effects described in #5176. So, don't attempt to clone when the destination already exists on these platforms.

• Resolves: Overwriting a file on a remote VFS deletes it without copying the new one (macOS, Solaris: file cloning) #5176

Checklist

• I have referenced the issue(s) resolved by this PR (if any)
• I have signed-off my contribution with git commit --amend -s
• Lint and unit tests pass locally with my changes (make indent && make check)
• I have added tests that prove my fix is effective or that my feature works
• I have added the necessary documentation (if appropriate)

[gereoffy]

Tested on macOS with an FTP destination (default settings, file cloning enabled): overwriting an existing remote file now does STOR with the new content, no DELE, so the data loss from #5176 is gone. Overwrites go through the normal O_TRUNC path again, which also keeps owner/permissions/hard links of local destinations. LGTM, thanks!

[zyv]

Thank you, folks!

[ossilator]

the commit message is (again) totally unacceptable by my standards. that makes it rather hard to argue about the change. i'll do that in the issue instead.

[zyv]

the commit message is (again) totally unacceptable by my standards. that makes it rather hard to argue about the change. i'll do that in the issue instead.

For me, a much more impressive revelation is that in the PRs that I have run through Claude so far, it seemed to always find a large number of data corruption and crasher issues. And these are not (only) pre-existing issues, of which there are more than enough, but the ones introduced in the new submissions. Also, this is putting all nitpicking aside - only genuine faults. None of that seems to be picked up by humans. Particularly saddening was the editor PR. On that backdrop, the commit message bickering just adds a layer of noise for me, unless it contributes to the crash suppression, which in theory it should, but in practice seems to not matter at all. At least this PR is more about saving the data than silently erasing the data, commit message notwithstanding.