small issues with midi plugout

[mmitch]

Checking out issue #19 I ran the midi plugout for the first time in a long time and found some small issues:

• [x] The plugout name in gbsplay -o list says MIDI sound driver but it is a file writer like stdout. This should be a simple rename.
• [x] The generated filename is gbsplay-1.mid. When you run it again, that file is silently overwritten. We should either output a warning, disallow the overwrite or sequentially name the output files (there is already a -1 in there, so using -2 and so on should not be too surprising to the user).
• [x] Silence detection does not work properly: When running gbsplay -o midi examples/nightmode.gbs, output stops after 2 seconds because gbsplay runs into the default silence timeout. -T has to be set to export more than 2 seconds. I'd guess silence detection does not work because no waveform is generated.

[mmitch]

D'oh: The number part of the filename is the subsong number, so we can't increment it for subsequent exports. The easiest way would be to add some documentation regarding filename creation and overwriting of existing files.

[mmitch]

task 1 fixed with e1872f6

[mmitch]

task 2 fixed with 729d1f8 (documentation only, no changes)

[mmitch]

@ranma I have tracked part 3 of this issue down to these two lines in gb_flush_buffer()

Two plugouts (iodumper and midi) don't implement the write callback because they don't need it. This enables the shoutcut highlighted above and gb_flush_buffer() returns eartly. Unfortunately, this also means that lminval, lmaxval, rminval and rmaxval don't get updated, which leads to the silence detection kicking in immediately in gbs_step()

In the midi output this also manifests as the global volume being empty in the status line during conversion, while the per-channel-volumes are displayed correctly.

What should we do?

1. remove the shortcut in gb_flush_buffer (this increases computation time during conversion - but who cares?)
2. add empty writer implementations to both plugins (but then the shortcut from 1. is never used and we can simply implement 1. straight away)
3. disable the silence timeout when using those two plugins (with a new flag like we did for stdout/verbosity)
4. add a note to the manpage that says you have to set -T or output will be cut off (I don't like this one)

What do you think?

[mmitch]

Thanks!