safe vs. unsafe FFI calls

[iustin]

The import of statvfs is:

foreign import ccall unsafe statvfs :: CString -> Ptr a -> IO CInt

Since this function is doing a system call, I'm not sure why it is marked as unsafe - in case of remote filesystems, this could block for a long time (in real world terms). What do you think about making this safe, since the overhead of the syscall will anyway make the Haskell RTS overhead most likely moot?

[redneb]

I don't feel strongly about either way. I was merely following the convention followed by other packages that system calls that do not generally block are imported as unsafe. Examples of that are unlink(2), rename(2), chown(2), stat(2) in the unix package. All of those function suffer from the same possible issue that you mention.

[iustin]

Interesting, thanks for the info - I'll followup there and then update this issue.

[iustin]

FYI, on haskell/unix#34, the agreement is that unsafe is the wrong approach. I'll try to send a pull request later this week (if you don't beat me to it).

[redneb]

I just uploaded a new version (v0.1.0.1) on hackage for that. Thanks for the report.