[PATCH 3 of 6] patch: migrate to util.iterfile

[Jun Wu]

Excerpts from Yuya Nishihara's message of 2016-11-15 21:14:10 +0900:
> On Mon, 14 Nov 2016 23:35:55 +0000, Jun Wu wrote:
> > # HG changeset patch
> > # User Jun Wu <quark at fb.com>
> > # Date 1479165246 0
> > #      Mon Nov 14 23:14:06 2016 +0000
> > # Node ID d1637df5ffd91e95da25213f06f346adb3269ace
> > # Parent  8097b8c66952003addd5b683a14265c1d3cc201f
> > # Available At https://bitbucket.org/quark-zju/hg-draft 
> > #              hg pull https://bitbucket.org/quark-zju/hg-draft  -r d1637df5ffd9
> > patch: migrate to util.iterfile
> 
> Not all read()s are interruptible by signals. I don't remember the detail, but
> signal(7) of Linux says "read(2) ... calls on "slow" devices. ... Note that a
> (local) disk is not a slow device according to this definition; I/O operations
> on disk devices are not interrupted by signals."

Good to know! I guess modern OS have similar decisions, although POSIX seems
to be not helpful here.

> Since readline() doesn't have cache in CPython layer, it would be slightly
> slower than fp.__iter__().

A quick benchmark shows it's ~4x slower.

Maybe make the function smarter to only use the workaround for a pipe?

  if hassafeattr(fp, 'fileno') and stat.S_ISFIFO(os.fstat(fp.fileno()).st_mode):
      return iter(fp.readline, '')
  else:
      return fp