Hi

Il 14/dic/2014 00:07 "Nikolay Dimitrov" picmaster@mail.bg ha scritto:

Hi Michael,

On 12/13/2014 04:32 PM, Michael Trimarchi wrote:

...

Hi

On Sat, Dec 13, 2014 at 10:16 PM, Eric Bénard eric@eukrea.com wrote:

...

Hi Nikolay,

Le Fri, 12 Dec 2014 20:16:19 +0200, picmaster@mail.bg a écrit :

Can you just change your enviroment? if I remember correctly you can override boot delay

Of course I can change the environment, but my point was towards improving the default "out-of-the-box" behavior.

I understand your point but I prefer a config option in configuration instead of change it. I'm not have this board but most of the time I always change something I don't like create my own environment or config.

Michael

Kind regards, Nikolay

On 13/12/14 17:03, Eric Bénard wrote:

Hi Nikolay,

Le Sat, 13 Dec 2014 18:05:15 +0200, Nikolay Dimitrov picmaster@mail.bg a écrit :

...

On 12/13/2014 04:16 PM, Eric Bénard wrote:

...

...

This will increase the boot time by 2 seconds : why is that needed ?

Thanks for commenting. My intent is to give by default a reasonable time for the end-user to react and break the boot-sequence if needed (and avoid the nasty racing against the clock for this precious key-press).

My motivation for this is that I see RIoTboard/MarsBoard mostly as boards that people can use for experimentation/prototyping, and not as a finished end-product, thus my thinking that the slightly increased boot-time won't be a critical issue for such applications. Still everyone can change the delay at will if they want to optimize the boot.

Other hobby boards also have similarly handy boot-timings, like Wandboard (5s), Novena (5s), Udoo (3s).

OK seems reasonable.

I disagree. I think it's an unnecessary, unneeded, unreasonable tripling of the boot time.

Rather than leaving people to patch out this unwanted behaviour, I'd suggest promoting CONFIG_BOOTDELAY into a Kconfig option, leaving the default as-is.

This allows the people who want it to configure it easily.

In any case the rationale given, that the Riotboard is a dev board used for experimenting, doesn't imply a change like this one. It implies a conversion to use config_distro_bootcmd.h like the patches I've just posted. This also leads to a slightly less unreasonable increase in bootdelay to 2s from config_distro_defaults.h

Using config_distro_bootcmd also means that we can largely ignore the u-boot shell altogether and just use extlinux.conf instead

Hi Iain,

On 12/14/2014 04:52 PM, Iain Paton wrote:

On 13/12/14 17:03, Eric Bénard wrote:

...

Hi Nikolay,

Le Sat, 13 Dec 2014 18:05:15 +0200, Nikolay Dimitrov picmaster@mail.bg a écrit :

...

On 12/13/2014 04:16 PM, Eric Bénard wrote:

...

...

...

This will increase the boot time by 2 seconds : why is that needed ?

Thanks for commenting. My intent is to give by default a reasonable time for the end-user to react and break the boot-sequence if needed (and avoid the nasty racing against the clock for this precious key-press).

My motivation for this is that I see RIoTboard/MarsBoard mostly as boards that people can use for experimentation/prototyping, and not as a finished end-product, thus my thinking that the slightly increased boot-time won't be a critical issue for such applications. Still everyone can change the delay at will if they want to optimize the boot.

Other hobby boards also have similarly handy boot-timings, like Wandboard (5s), Novena (5s), Udoo (3s).

OK seems reasonable.

I disagree. I think it's an unnecessary, unneeded, unreasonable tripling of the boot time.

I have 2 things to say about this: 1. In theory we can also question why the initial boot delay was selected to be 1s. 2. You imply that the "users" want ultra-fast boot times for riotboard. Being myself one of the riotboard users, I can assure you that such generalization is excessive.

Rather than leaving people to patch out this unwanted behaviour, I'd suggest promoting CONFIG_BOOTDELAY into a Kconfig option, leaving the default as-is.

This is a good idea in general, I don't mind it.

This also leads to a slightly less unreasonable increase in bootdelay to 2s from config_distro_defaults.h

Well, this is again subjective - we can always argue that 2s is both too long and too short for selected use cases, so I don't understand the qualification "less unreasonable".

In summary - I won't carry a war for this change, as it doesn't worth it. I just tried to contribute a tiny bit of improvement to the already excellent work of Eric, so we can have a silk-smooth out-of-the-box U-Boot user experience, and I'm doing this as an actual riotboard user. That's all.

Thanks for sharing your viewpoint.

Kind regards, Nikolay

On 14/12/14 20:56, Nikolay Dimitrov wrote:

In summary - I won't carry a war for this change, as it doesn't worth it. I just tried to contribute a tiny bit of improvement to the already excellent work of Eric, so we can have a silk-smooth out-of-the-box U-Boot user experience, and I'm doing this as an actual riotboard user. That's all.

I'm a riotboard user too, and have contributed towards it's success in my own ways. The choice of 1s hasn't been a problem for the many people using the Debian images I provided.

Anyway, it's in the interest of Riot/Mars to move towards the config_distro_* way of doing things longer term as it's likely to gain easy support by the general distros that way.

Assuming those patches get accepted, the bootdelay ends up in the common config shared by all boards using config_distro_defaults. Over time I suspect that's likely to include most of the general dev boards.

It's still likely to be worth moving it to Kconfig as both of us will probably want to change it to our own preferences regardless :)