Jeremy White wrote:
The attached patch fixes all remaining test failures in winmm when run in interactive mode on my box with Wineoss.
It does so by making msacm conversions start aligned (which lets them finish aligned for my test cases).
However, it's clear that Robert Reif intended them not to be aligned. My question is: why?
The reason it is that way is because it duplicates windows behavior exactly. Run the same tests with a windows msacm32.dll any you get the exact same errors.
The problem is that wine's wave mapper expects msacm to do the right thing but microsoft's msacm doesn't. Microsoft states that you must check the amount of data converted because there is no guarantee that it will be the amount requested. You may need to do another conversion on the remaining data to get it all. Thats what the wave mapper needs to do to get all the data.