While your points are good and valid, I'd like to point out there is is a good reason to invoke cp via system instead of using File::Copy: The latter doesn't preserve file permissions on Unix and Linux.
I personally believe this is very well known, with Abigail often advocating its use. Another issue I've "always" been wondering about has to do with sparse files: if I create one and then copy it with cp, IIRC the target file will also be sparse and a quick experiment now confirms my memory:
1;0 blazar@feather:~/tmp$ perl -e 'open F, ">foo.txt"; seek F, 2**30, +0; print F 1' 1;0 blazar@feather:~/tmp$ ls -lh . ; du -sh . total 12K -rw-r--r-- 1 blazar blazar 1.1G Nov 16 10:22 foo.txt 16K . 1;0 blazar@feather:~/tmp$ time cp foo.txt bar.txt && echo -e "\a" real 0m2.357s user 0m0.204s sys 0m2.064s 1;0 blazar@feather:~/tmp$ ls -lh . ; du -sh . total 24K -rw-r--r-- 1 blazar blazar 1.1G Nov 16 10:22 bar.txt -rw-r--r-- 1 blazar blazar 1.1G Nov 16 10:22 foo.txt 28K .
It takes some time (which further increases with the size of the file) and I don't have the slightest idea of how it manages to do so... (perhaps some fiddling with strace(1) may help, but I should be more confident with it.) But it succeeds. How does File::Copy behave in this respect? I believe it will create a file with an actual disk usage comparable to its own size, filled with nulls, ain't it?
Granted, I would regard this as a minor but not completely irrelevant issue.