Helmut Karlowski wrote:
It saves and restores the CPU-regs. There were severe problems with USERDEFs before.
I double checked that, there is currently no register clobbering problem in my testcase.
However, currently EmuTOS calls USERDEFs' callbacks as a simple function call ! So, when EmuTOS and the CF-Lib are both compiled with the same compiler (currently GCC for me), there is no trouble. But problems may appear if the CF-Lib is compiled with a compiler which clobbers more registers than GCC.
What is the rule about USERDEFs callbacks ? Which registers can be clobbered ? Which ones must be preserved ?
So VDI_INTINMAX became larger?
Good question.Maybe the old QED/CF-Lib binary was compiled with Pure C, which had different GEM bindings ?
-- Vincent Rivière