"Alexander E. Patrakov" patrakov@gmail.com wrote:
Andrew Eikum converted the buffer position calculation code from fixed point to floating point, because floating-point code is more readable. I agree with him.
I am not sending further patches out, because it is not yet clear which of the two proposed patchsets for bug 14717 will be the final one. Anyway, both of the approaches depend on all the patches that I sent today.
Please ignore this patch for now, it has at least one bug.
- freqAcc += adj; - adv = (freqAcc >> DSOUND_FREQSHIFT); - freqAcc &= (1 << DSOUND_FREQSHIFT) - 1; - ipos += adv * istride; + freqAcc += dsb->freqAdjust; + ipos += ((DWORD)freqAcc) * istride; + freqAcc -= truncf(freqAcc);
The last line of this hunk is wrong, should be "=", not "-=".