Re: [PATCH 1/1] Improve progress display in kB range.

James Cloos [off-list ref] writes:

|> Therefore, in practice — and as I have witnessed several thousand times
|> without ever having seen a contrary example — display_throughput() is
|> called *durring* a download only when total & 0xFFF == 0xFFF.

Another possibility is an off-by-one error.  The relevant part of fill()
looks like:

,----< excerpt from index-pack.c:fill() >
|   do {
|     ssize_t ret = xread(input_fd, input_buffer + input_len,
|                         sizeof(input_buffer) - input_len);
|     if (ret <= 0) {
|       if (!ret)
|         die("early EOF");
|       die("read error on input: %s", strerror(errno));
|     }
|     input_len += ret;
|     if (from_stdin)
|       display_throughput(progress, consumed_bytes + input_len);
|   } while (input_len < min);
|   return input_buffer;
| }
`----

if *(input_buffer + ret) is the last read octet rather than the next
empty octet, that would also explain what I see.

After checking "ret" from xread(), input_len is incremented by that
amount, and the next iteration gives "input_buffer + input_len" to
xread().  If input_buffer[ret] _were_ the last octet read, your loop would
be discarding that octet when you call more than one xread() to fill the
buffer.