Update sifcmd.c

[psxdev]

Fixed support for user land ee sifcmd registering , it was lost i don't know why.

[sp193]

You were supposed to call SifSetCmdBuffer(SifCmdHandlerData_t) first.

[rickgaiser]

Cross referencing usr_cmd_handlers or nr_usr_handlers would have been enough to see these values already get used, and shows how a user application can/should use them:

ps2sdk/ee/kernel/src/sifcmd.c

Lines 305 to 314 in e8e9edb

	SifCmdHandlerData_t *SifSetCmdBuffer(SifCmdHandlerData_t *db, int size)
	{
	SifCmdHandlerData_t *old;
	old = _sif_cmd_data.usr_cmd_handlers;
	_sif_cmd_data.usr_cmd_handlers = db;
	_sif_cmd_data.nr_usr_handlers = size;
	return old;
	}

Please leave changes like these open a little longer next time so others can comment before they are committed. 1 or 2 days would be enough I think.

I think we should revert this commit.

[fjtrujy]

Ok @psxdev can you try the solution proposed over here, if this is working, I will revert the change.

Thanks & sorry

[sp193]

If I remember right, we originally had a static buffer of a fixed number of entries, hence the OP's question. I probably changed it to fit the official behaviour because it was undocumented; since the handler buffer size is fixed, one could only use SIFCMD IDs up to the last index, but yet this number is not documented. It was neither possible to reduce its size nor to increase it.

[psxdev]

you can revert the change however if SifAddCmdHandler is called before set the user buffer it will crash without message

[sp193]

That's because the documentation wasn't followed. At least it's possible to get the exception trapped. If we want to make this idiot-proof in the sense that it won't crash, then SifAddCmdHandler should probably be made to do nothing - but then it'll become a silent fail. If we printed an error, then this becomes coupled with some library.