[physfs] [semi-OT] stdio question

[Ryan C. Gordon]

> My assumption is that the seek is deferred until an actual file 
> operation, and that it just updates a cursor internally.  At least, I 
> would seriously hope that's how most implementations work.

I wouldn't assume that...it's likely that many runtimes do a one-to-one
mapping with the system-level API and don't optimize like this...in the
case of seeking, the assumption is frequently "it's file i/o, don't do
it in time-critical code anyhow".

Doubly so because almost every seek is immediately followed by a
read/write, so it's less worth delaying the seek.

That being said, seeking to the end followed by an ftell() is probably
more portable, but you can also use fileno() to get a handle from a
(FILE *) to use with fstat()...on supported platforms, that is, most
platforms, this is probably faster.

--ryan.