Moved 2 mirrored drives with SmartOS install from one to another physical machine and got this on boot (BIOS, UEFI boot doesn't work) mega.nz/file/sBMVgTxB#hPE5c8hyz07DCsaTQ18qXxomZtrCF_fi1vM4lTyAuE4

Catch is that on physical serial port on that machine, (HP 600 G1 TWR), that is turned on serial modem (one that can communicate over POTS lines).

If serial modem is turned off, such (flashing and then displaying) message is not displayed.

Could be loader or Smartos defaults are expecting serial console on serial port and not expecting other devices could be connected to serial ports.

e.g. it won't boot..

if you connect something on serial port, like modem.

nikolam: Loader is configured to have a console on serial ports 1-4. SmartOS defaults to ttyb as the system console, but will spawn non-system consoles on all serial ports ttya-ttyd post-boot.

nikolam: If you have a device that's spewing characters at the serial console, then it's going to react.

That's considered the correct behavior.

So it won't boot then..? could that be changed in loader settings.. what serial ports to leave alone..

I also have a PC (triton headnode install) where , if monitor (DP to HDMI) is connected then physical console is OK. If monitor is not connected during the boot, physical console displays up to loader messages but does not work post-boot.

And screen is i some low res

Similar thing on same type of machine is also on Linux desktop distro, if monitor is not connected during boot, if never starts normal X session after connecting monitor.

loader will run consoles on the first 4 serial ports. On the boot media you can modify loader.conf to change that.

BIOS loader will pick up serial port set via ACPI SPCR (console redirection to serial in BIOS setup), and therefore you can comment out console line from /boot/defaults/loader.conf.

UEF loader is currently buggy and needs this help

I intend to fix it to pick up devices from ConOut variable, but this work will get some time yet (and there is huge bear trap)

Maybe could continue booting whatever is on serial console.. probably interprets anything sent over serial console as terminal and waits forever

and at the same time, VGA/console display dies displaying serial data..

so KVm over IP won't help you, only booting from USB and selecting to disable some of serial lines in loader config. And if you don't know that is the cause.. won't ever boot.

Maybe if loader continues to count up to 60 seconds counter, even with user activity and boots. unless user explicitly confirms to stop counter in a new loader option?

And if it runs conseoles on serial ports, why also not running console on VGA/display output at the same time.

It would need VT service then.. because serial console is default.. ok

Btw, I mentioned earlier I think, seems that displaying 'man' pages on graphical console displays garbled characters..

nikolam which ones?

and which font is used?

tsoome, I think it is whenever text in man in bolded..

looks ok for me:)

this one is console from UTM (qemu)

800x600 pixels, 8x16 terminus

nikolam: What's happening is that you have a device sending characters to the serial console. The only proper thing for loader to do is interpret that as a user pressing keys. Ignoring that input means that we would also have to ignore user input from a keyboard.

We can't tell what's on the other side of the serial line. All we know is we received data that is indistinguishable from keyboard input.

dladm: add operation failed: I/O error

teutat3s: should get us a nice void linux hvm image

bahamat, well yes, so keybpard input can have a timeout , if nothing pressed in 60 seconds, continue to bot.

tsoome, i5 4570, intel VGA out, HP 600 g1 twr, man man , Smartos GZ mega.nz/file/RQdyHKJQ#YLYE4ROHfhuVen_jqsWuR7D2yHjKYIQ8bFEFoNKY9nk

hm, thats underline gone bad

interesting

oh, what font are you using there?

Any chance `man` is doing the dance where it emits scritch backspace character scritch backspace character lather rinse repeat?

in your case, it is not only underline having issues, but also -

'cuz it looks like _ is rendering as a box

no, not man. the console text renderer does query font for glyph, and if there is none, the "default" glyph is used. Underline is printed by using 2 glyphs - the symbol itself and then glyph fo underline.

the translation is usually done by the pager, though

the "default" for missing glyph is rectangular box, so you get symbol + box in the same glyph space.

basically, the problem is that in this case the font seems to be missing unicode chars

in our case, font_bit_to_pixX() is using 0x0332 (combining low line) to make the underline effect.

oic

see usr/src/common/font/font.c

the terminus font we are using by default, has 0x0332 added (original ones are missing it)

I also did add it to /boot/fonts/Gallant19.fnt, but since it has a lot less glyphs compared to terminus, I did not add Gallant to font index file, so it wont be used automatically (but one can use it with loadfont command from loader)

tsoome, never were editing fonts usage, I guess it is default..

in that case, it is built in default, which is limited to ascii set.

you can test it with iconv -f 8859-1 -t UTF-8 /usr/pub/iso

you will probably see a0-ff as those rectangular boxes.

I guess, it would be time to implement font loading from userspace;)

may need iso-8859-1, and apparently it doesn't work in a gz?

iconv is present, but seems to be missing some of its data

/usr/lib/iconv/geniconvtbl/binarytables: No such file or directory

triton gz says: Not supported 8859-1 to UTF-8 , Smartos gz says the sam..

CP819 IBM819 ISO-8859-1 ISO-IR-100 ISO8859-1 ISO_8859-1 ISO_8859-1:1987 L1 LATIN1 CSISOLATIN1

Those are the valid names, if iconv has its needed data files.

from `iconv -l`

ah, right. you can convert it in local zone then and cat it to console.