Re: [Swsusp-devel] Re: swsusp problems [was Re: Your opinion on the merge?]

["Michael Frank" <mhf@linuxmail.org>]

On Thu, 25 Mar 2004 00:23:38 +0100, Pavel Machek <pavel@ucw.cz> wrote:

> On Čt 25-03-04 06:46:12, Michael Frank wrote:
>> May I request that you leave the authors headers intact when quoting. Thank
>> you.
>
> As you wish.
>
>> On Wed, 24 Mar 2004 11:17:04 +0100, Pavel Machek <pavel@ucw.cz> wrote:
>>
>> >>>>So why aren't you arguing against bootsplash too? That definitely
>> >>>>obscures such an error :> Of course we could argue that such an error
>> >>>>shouldn't happen and/or will be obvious via other means (assuming it
>> >>>>indicates hardware failure).
>> >>>
>> >>>Of course I *am* against bootsplash. Unfortunately I've probably lost
>> >>>that war already. But at least it is not in -linus tree (and that's
>> >>>what I use anyway) => I gave up with bootsplash-equivalents, as long
>> >>>as they don't come to linus.
>> >>>
>> >>>[And I believe Linus would shoot down bootsplash-like code, anyway.]
>>
>> Why? Joe consumer wants it.
>> As to the ever growing size of the kernel, there could be a official
>> addons/tools
>> tree with non-core functions maintained by a seperate maintainer. Things
>> like
>> debuggers, profiling or (swsusp) debug support could go there as
>> well...
>
> Yes, having -nice patch with bootsplashes, translated kernel messages,
> and swsusp eye-candy would work for me.

If a -nice _tree_ is useful, your guys just have to launch it. Gosh this could reduce
arguments about what goes into the kernel and save Linus and Andrew lots of work.

> Feel free to maintain it.

Busy enough with testing, actually far too busy for being on a volunteer basis ;)

I am sure that better qualified and properly supported/sponsored individuals
will queue up as long as it is an _official_ -nice tree with the good purpose
of centralizing useful non-core functions :)

>
>> >>Solution: Auto switch to non-swsusp VT on error showing the error message.
>> >
>> >Hmm, at that point you loose context, like now you know what error
>> >happened, but do not know at which phase of suspend. That's pretty bad
>> >too.
>>
>> Right, Good idea!  Just  print always "ugly" swsusp context on a text VT -
>> plus any
>> error messages  - and switch over to this VT in printk when not in interrupt
>> context. 10 lines of code or so in printk ;)
>
> You see, 10 lines in printk is probably good enough reason not to
> include that patch in kernel, because its "too ugly".

Pretty does not count above, Ugly does not count here, Functionality always does.
Besides that patch might be in the -nice tree.

> Plus it does not work if printk _was_ from interrupt context.

Kernel knows when in interrupt context and can defer switching.

>
> swsusp really should not have patch any code outside kernel/power.

Which is extremely ideal, but one thing at the time...

Michael