[Media_api] Windows midiOutLongMsg strangeness

[Ben Allison]

Roger,

Just to make sure we're on the same page, here's what the revised code 
would look like with the prepare/unprepare stuff in place

        m->hdr->dwBufferLength = m->sysex_byte_count;

        /* DEBUG CODE: */
        { int i; int len = m->hdr->dwBufferLength;

          /* added */
          if (m->hdr->dwFlags & MHDR_PREPARED)
             midiOutUnprepareHeader(m->handle.out, m->hdr, sizeof(MIDIHDR));
          /* ----- */

          char *newmsg = (char *) malloc(len);
          for (i = 0; i < len; i++) newmsg[i] = m->hdr->lpData[i];
          m->hdr->lpData = newmsg;

          /* added */
          m->hdr->dwFlags=0;
          midiOutPrepareHeader(m->handle.out, m->hdr, sizeof(MIDIHDR));
          /* ----- */
        }

        m->error = midiOutLongMsg(m->handle.out, m->hdr, sizeof(MIDIHDR));
        if (m->error) rslt = pmHostError;

-Ben Allison

Roger Dannenberg wrote:
> Yes, I call unprepare when buffers are returned, but in most of these 
> tests, I'm using the buffer for the first time when the failure 
> occurs. I'm checking that midiOutPrepareHeader sets the dwFlags field 
> like it's supposed to. (Keep those ideas coming -- there must be 
> something I haven' thought of. I've asked the Midi Ox developer about 
> some internal details of their code, and I've checked rtMidi -- which 
> uses malloc() for every sysex send -- and I'd be interested to see any 
> other source that successfully sends sysex messages.) -Roger
>
> ----- in reply to ------
>
> Roger,
> Did you unprepare (midiOutUnprepareHeader) first?  If the MIDIHDR that 
> was
> passed in was already prepared, you'll need to do this.  The MIDIHDR data
> structure stores a bool to indicate whether or not it's already been
> prepared.  If it's been prepared already, midiOutPrepareHeader exits
> immediately without doing anything (or at least the documentation 
> suggests
> so).
>
> -Ben Allison