Re: console in 11.0-ALPHA4

John Baldwin wrote:
> On Monday, June 20, 2016 04:54:11 PM Ernie Luzar wrote:
>> Ed Maste wrote:
>>> On 20 June 2016 at 14:29, Ernie Luzar <luzar722_at_gmail.com> wrote:
>>>> I found the cause of this boot time message
>>>> "vicontrol: setting cursor type: Inappropriate ioctl for device"
>>>>
>>>> In my rc.conf I had this statement
>>>> vidcontrol -c blink -h 250
>>>> From testing it seems that vt does not handle the "blink" option for causing
>>>> the cursor to blink. Is there a vt option to enable cursor  blinking
>>> I've submitted two PRs for the issues you reported here:
>>> Bug 210413 - vidcontrol -c <normal|blink|destructive> does not work with vt(4)
>>> Bug 210415 - vidcontrol -h <history size> does not work with vt(4) (edit)
>>>
>>> There is currently no option to enable cursor blinking.
>>
>> Further testing shows that:
>>
>> 1. The rc.conf option screen saver "saver= " and the "blanktime= " 
>> options work in vt for both text and graph modes.
>>
>> 2. The cursor copy/paste does not work in vt text mode. It only works in 
>> vt graph mode. vt needs to be fixed so copy/paste function also works in 
>> text mode.
>>
>> 3. Boot time splash screen does not work in vt at all no matter which 
>> mode is enabled. Using this in loader.conf
>> splash_bmp_load="YES"
>> bitmap_name="/boot/splash.bmp"
>> bitload_load="YES"
>>
>> This works in 10.x. The splash screen function can not be allowed to be 
>> removed from the base system just because vt can not handle it. This 
>> also means any vt changes need to updated in the handbook section on 
>> splash screen.
>>
>> In conclusion, vt is not ready to replace the sc console driver yet. vt 
>> needs to be fixed before becoming the default in 11.0. Using vt as the 
>> default console driver effects every user. It completely changes the 
>> FreeBSD experience. It's detrimental to the public relations and the 
>> professional image of FreeBSD to publish the 11.0-RELEASE using vt in 
>> its current condition. If time doesn't permit to get these vt things 
>> fixed before publishing 11.0 then vt should not become the default 
>> console driver.
> 
> OTOH, if you use EFI, then you can't use sc(4).  You get no working console
> with EFI (which is becoming more popular) if you use sc(4).  You also do
> not get non-ASCII characters with sc(4), so you don't have a UTF-8 capable
> console if you use sc(4).  You also do not have a working console after
> loading the KMS drivers (either by starting X or via explicit kldload) when
> using sc(4).  This affects just about every modern Intel system using an
> Intel GPU (so more widespread than EFI even).
> 
> There are trade offs in both directions.  Neither console is a subset of the
> other.  However, sc(4) is not really expendable to support the things it is
> missing.  vt(4) is actively worked on, and patches for the features it lacks
> that you need are certainly welcomed.
> 

This sounds like a integration design error. From your description of 
the hardware that vt is designed for and sc is designed for indicates 
the need for some code to be added to the boot process that inspects the 
hardware and decides whether vt or sc is to be launched. Or maybe 
bsdinstall should inspect the hardware and give the installer the option 
to select what console is best for his hardware. Just forcing this 
version of vt that lacks long time functions as the default console 
driver is not the professional way to handle such conflicts.

I am at a lost to understand how any development administrator would 
approve making vt the default console driver in light of the standard 
functions missing from it. And furthermore what kind of vt testing was 
done that these problems were missed. Any one of these problems is 
enough cause to reverse the decision to make vt the default console 
driver for 11.0. This would give time for these problems to be address 
and correctly tested and when ready then be paced in some other 11.x 
release.