- proftpd should not rely fsio I/O is instant

[Sashan]

one of our customers hit this glitch on back up server, which serves files from SAM-QFS [1] volume. The I/O on such FS can take up to several seconds. This typically happens when file must be retrieved from tape.

In this case the FSIO read(2) operation fails with EINTR error. The EINTR failure is mishandled further: instead of reporting error to remote client, server sends file with 0 size.

The patch also makes the write(2) more robust. No one his hit such issue so far, but I believe similar failures might happen when received files are written to NFS file systems.

the issue got uncovered during upgrade of 1.3.5b to 1.3.6. Unlike 1.3.5b the new 1.3.6 guarantees a minimum alarm(2) interval to be a 5sec.

[1] https://en.wikipedia.org/wiki/QFS

[coveralls]

Coverage remained the same at ?% when pulling 1e29b1168b154e6e096f3a863a7198831a67bcdd on Sashan:fsio.eintr into 1e75710d43edb07aea34a200c839d67de698bba1 on proftpd:master.

[Castaglia]

I believe that now, with https://github.com/proftpd/proftpd/pull/1359, these occurrences should be addressed. Thanks!