Re: [OpenOCD-devel] sam3.cpu: IR capture error; saw 0x0f not 0x01

[Freddie C. <fre...@op...>]

W dniu 2012-12-05 08:44, Andreas Fritiofson pisze:
> You may be right, but then it sounds like a misfeature. Why shouldn't a
> target specific behavior override the generic mechanism?

It was like this when this cortex-m3 option was introduced, but 
fortunately it was changed, because then all targets used soft resets 
even if h/w reset was available, and that last one is always better 
(especially for targets where sysrstreq is not working and vectreset 
doesn't reset peripherals - like LPC17xx where you have to explicitly 
state the core clock to flash it). If user specifies that there is a 
reset than this reset should be used. cortex-m3 soft reset is defined in 
cortex-m3 target configs and there's really no point in editing this 
every time you change something...

You'll probably be able to easily find the discussion about that thing 
back then - I think it all goes down to not-very-lucky design of OpenOCD 
where everything is imperative, while it would be better if it would 
work like that:
- interface specifies what it supports (what reset signals it has)
- target/board specifies what it supports (what resets are supported, 
what lines are routed and how, ...)
- OpenOCD chooses the best possible combination (interface has 2 resets 
but target has only 1? use "srst_only" then)
- user can optionally override that

With current design (almost) no interface and no target config specifies 
resets as this wouldn't work as expected - users always have to specify 
them by hand and usability suffers... No one will change the fact that 
the most "user friendly" software is the one that doesn't need much 
configuration and OpenOCD is not among such programs.

4\/3!!