[flac-dev] Disk fragmentation

[lvqcl]

Martijn van Beurden <mvanb1 at gmail.com> писал(а) в своём письме Wed, 24 Sep 2014 16:09:38 +0400:

> Well, from the code it's clear that quite some effort was put
> into making libFLAC usable on platforms that have no FPU, as
> there is a integer-only define that actually works. I could
> think of small handheld recorders like the Zoom H1 or something
> like that to use FLAC.
>
> On the other hand, devices with no FPU are usually embedded, and
> embedded developers change some code anyway. Maybe it's a good
> idea to just mention this in the README, under the "Note to
> embedded developers", as something that can be removed without
> problem?

IMHO it's better to move setvbuf() out of libFLAC/stream_encoder.c
into flac/encode.c.  I can see two ways to do this:

1) Add a new function to libFLAC API that have access to
encoder->private_->file and can call setvbuf:

FLAC_API FLAC__bool FLAC__stream_encoder_set_buffer_size(FLAC__StreamEncoder *encoder, unsigned size);


2) Rewrite EncoderSession_init_encoder() in flac/encode.c so that it
opens the file, calls setvbuf and then uses FLAC__stream_encoder_init[_ogg]_stream()
instead of FLAC__stream_encoder_init[_ogg]_file(). It's more difficult,
but doesn't require API change.