On Sunday 17 May 2009, Rick Altherr wrote:
> Rather than combine them, I'd like to see jtag_queue_command() enforce  
> validation of the command to be enqueued.  Then the patterns would be:
> 
> cmd = cmd_queue_alloc();
> 
> cmd->type = JTAG_SCAN;
> ....

Then how about passing JTAG_* to the allocator?

Related approach:  command-specific allocators, which take
the parameters that will be fed to each command.


> 
> jtag_queue_command(cmd);
> 
> Otherwise, you place a command in the queue before it is filled out.

Yeah, that bothered me a bit too, but it *is* the current idiom.

Some jtag_queue_scan(...) calls -- allocate and set up each flavor
of command, then queue it -- could address that issue *and* keep
to the current idiom.


> This works fine today where the JTAG queue is manually flushed, but if  
> we ever went to a opportunistic queue draining scheme (device driver  
> pulls next item off queue automatically when a command finishes), you  
> could get bogus commands being executed.
_______________________________________________
Openocd-development mailing list
Openocd-development@lists.berlios.de
https://lists.berlios.de/mailman/listinfo/openocd-development

Reply via email to