Re: Tiny Core v17.0 upgrade issues

[Juanito]

I'm not certain of this and may be misremembering.
I think the new kernel is built on the previous version of Tinycore.
If that's the case, it might be built with the previous version of GCC.
Newer versions of GCC typically build larger executables.

Maybe Juanito and/or Paul_123 can comment.

The new toolchain is built in a chroot environment against the new kernel headers on the previous version of tinycore and this toolchain is used to build the new kernel. Thus the version of gcc in tc-17.x was used to build the kernel.

[Paul_123]

There were quite a few changes in the kernel config between 17.0 and 17.1 proposed kernel, in addition the kernel version.  That is why its size changed.

It should not make a difference what CPU a kernel is compiled under, as the kernel config dictates the compiler output, as long as its the same compiler version.   It is true that the x86 kernel is built on a 64bit cpu, but done in the tc 32bit environment.

[Stefann]

There were quite a few changes in the kernel config between 17.0 and 17.1 proposed kernel, in addition the kernel version.  That is why its size changed.

It should not make a difference what CPU a kernel is compiled under, as the kernel config dictates the compiler output, as long as its the same compiler version.   It is true that the x86 kernel is built on a 64bit cpu, but done in the tc 32bit environment.

So.. that is strange. Because on TC17.0 I compiled a TC17.1 kernel that is 20kB bigger than the published TC17.1 kernel.
It’s probably handier to investigate this towards TC17.1 instead of TC17.0 as the TC team has that under active development.

I’m currently out.
I will do a “very clean” TC17.1  kernel make tomorrow.
Including Mrproper and oldconfig, and make sure to have all apps update before I compile.

I wil do that TC17.1 compile on TC17.0 system. That’s ok I hope.

Anything else I should check?

[Stefann]

update.......
system was still running after 4 full days. I stopped it manually
So: "self compiled kernel with integrated usb-serial driver" WORKS
I tested with "no reads from usb serial" in my main application and have 200x accelerated 2byte block reads in dedicated testprogram running in parallel:
mon day:  5 23 | hh:mm:ss: 08:52:36 | Still alive after 0k + 1 loops
....
mon day:  5 27 | hh:mm:ss: 06:22:00 | Still alive after 130912k + 800 loops

So... strange....
- vmlinuz from TC17.0 download location including usb-serial-6.18.2-tinycore.tcz crashed 2x in a row. After 14hr and after 12hr.
- self compiled vmlinuz with integrated usb-serial survives 4 days. 8x longer.... no crash but manually stopped

To explore further:
I started self compiled kernel WITHOUT integrated usb-serial together with  usb-serial-6.18.2-tinycore.tcz
We'll see........
If this is "bad" I expect it to fail in 14-ish hours.

=====
Update on filesize difference of bzImage between published file versus self made file.
I generated a full fresh "from scratch" image for TC17.2
All done on hp510 2core 1GHz 64bit cpu running on x86 TC17.0

What i did:   
 

Code: [Select]

Check that I run latest TC17.0 vmlinuz:
TC17.0 download location:
vmlinuz                                            10-Feb-2026 12:03             6087168

tc@hp510:/mnt/sda1/tce/boot$ ls -l *17
-rwxrw-r--    1 tc       staff      6087168 Feb 10 13:10 vmlinuz17

Download may-8 version of TC17.1 from http://tinycorelinux.net/17.x/x86/release/src/kernel/
linux-6.18.28-patched.tar.xz                       08-May-2026 15:47           159508120
config-6.18.28-tinycore                            08-May-2026 18:09              218173

On hp510 system:
tc@hp510:/krubo/work/TC$ tar xvf linux-6.18.28-patched.tar.xz
tc@hp510:/krubo/work/TC$ cd *28
tc@hp510:/krubo/work/TC/linux-6.18.28$ cp ../config-6.18.28-tinycore ./.config
tc@hp510:/krubo/work/TC/linux-6.18.28$ make mrproper
  CLEAN   .config
tc@hp510:/krubo/work/TC/linux-6.18.28$ cp ../config-6.18.28-tinycore ./.config
tc@hp510:/krubo/work/TC/linux-6.18.28$ make oldconfig
  HOSTCC  scripts/basic/fixdep
  HOSTCC  scripts/kconfig/conf.o
  HOSTCC  scripts/kconfig/confdata.o
  HOSTCC  scripts/kconfig/expr.o
  LEX     scripts/kconfig/lexer.lex.c
  YACC    scripts/kconfig/parser.tab.[ch]
  HOSTCC  scripts/kconfig/lexer.lex.o
  HOSTCC  scripts/kconfig/menu.o
  HOSTCC  scripts/kconfig/parser.tab.o
  HOSTCC  scripts/kconfig/preprocess.o
  HOSTCC  scripts/kconfig/symbol.o
  HOSTCC  scripts/kconfig/util.o
  HOSTLD  scripts/kconfig/conf
#
# No change to .config
#
tc@hp510:/krubo/work/TC/linux-6.18.28$ make bzImage
tc@hp510:/krubo/work/TC/linux-6.18.28$ cd arch/x86/boot
tc@hp510:/krubo/work/TC/linux-6.18.28/arch/x86/boot$ ls -l bz*
-rw-r--r--    1 tc       staff      6312448 May 27 12:42 bzImage

bzImage on TC17.1 download location http://tinycorelinux.net/17.x/x86/release_candidates/tc17.1/
bzImage                                            08-May-2026 18:33             6111744
So....
My bzImage is 20k bigger than the published version while published image has may-8 timestamp and sources I used also have may-8 timestamp.

It's probably simply something "configuration" and nothing problematic

For now conclusion & way forward is:
- self compiled TC17.0 kernel works while published 17.0 kernel crashes on usb-serial read.
- I will dig a bit deeper into it but at some time "just start using the self compiled kernel". There is no real downside on that.
- Once TC17.1-beta gets released I will test published image again. Maybe the version update will have fixed things.

[Rich]

Hi Stefann
Just a few minor observations.

... What i did:   
 

Code: [Select]

----- Snip -----
tc@hp510:/krubo/work/TC/linux-6.18.28$ cp ../config-6.18.28-tinycore ./.config
tc@hp510:/krubo/work/TC/linux-6.18.28$ make mrproper
  CLEAN   .config
tc@hp510:/krubo/work/TC/linux-6.18.28$ cp ../config-6.18.28-tinycore ./.config
 ----- Snip ----- ...

Now you see why you run  make mrproper  before copying the config file.

A quick lesson in paths and syntax.
./  Means the current directory.
../  Means the directory right above the current one.

If you want to run a program that's located in the directory you are in:

Code: [Select]

./ProgramNameThis tells the shell to look in the current directory instead of searching $PATH:

Code: [Select]

tc@E310:~$ echo $PATH
/home/tc/.local/bin:/usr/local/sbin:/usr/local/bin:/apps/bin:/usr/sbin:/usr/bin:/sbin:/bin:/etc/sysconfig/tcedir/ondemand

Code: [Select]

cd ./        # This does nothing
cd ../       # This moves you up one level in the directory

Code: [Select]

tc@hp510:/krubo/work/TC/linux-6.18.28$ cp ../config-6.18.28-tinycore ./.config        # The ./ is redundant
tc@hp510:/krubo/work/TC/linux-6.18.28$ cp ../config-6.18.28-tinycore .config          # This accomplishes the same thing
None of the above are a problem, I just wanted to offer some clarification.

This, on the other hand:

...

Code: [Select]

----- Snip -----
tc@hp510:/krubo/work/TC/linux-6.18.28/arch/x86/boot$ ls -l bz*
-rw-r--r--    1 tc       staff      6312448 May 27 12:42 bzImage

bzImage on TC17.1 download location http://tinycorelinux.net/17.x/x86/release_candidates/tc17.1/
bzImage                                            08-May-2026 18:33             6111744
So....
My bzImage is 20k bigger than the published version  ...

makes me question your math skills.

Code: [Select]

tc@E310:~$ calc 6312448-6111744
200704
While I routinely do math in my head, I do tend to check my result
by using calc which is available on the command line.

[Stefann]

@Rich
Ah... yes indeed... mrproper gets rid of the .config....

Yeah... I know the meaning of ./ and ../
I just wanted it to make it "Cristal clear" to show .config is really .config and not ./config

Having said that...
OUCH....
2nd time you catch me on a 10x error!
(and as I said earlier, I really appreciate your keen eye for details. If we ever understand this issue I'm pretty sure it's a damn detail)

200kB files difference is "really something".

Anyway....
We'll see how it goes I guess.

mmmmh... calc is a fun tool. Thanks

[Rich]

Hi Stefann

... mmmmh... calc is a fun tool. Thanks

Yes it is. Just bear in mind, it's a simple calculator:

Code: [Select]

tc@E310:~$ cat /usr/bin/calc
#!/bin/busybox ash
. /etc/init.d/tc-functions
useBusybox
if [ -z ${1} ]; then
  echo
  echo "         -=[ Simple Awk Calculator ]=-"
  echo "Put formula in quotes (single or double) examples:"
  echo "calc '3*4'"
  echo "calc 'sqrt(9)'"
  exit 0
fi
awk "BEGIN{ print $* }"

[Stefann]

update...
This morning all was still running.
However.....
I had a major breakdown of my 1-wire control network yesterday evening.
This is completely unrelated, it looks like a hardware breakdown of the "hub" which brings the total network down.
So... current configuration had 2x 12hr testtime instead of 1x 24hr
With an "expected time to crash in case of buggy" of 12..14hr this test cannot be seen as conclusive.

I will be one a trip until Sunday. During that time I run all on TC15 to avoid issues.
Sunday onwards I will have to divide my time over 1-wire hardware problems and TC17 exploration.

Summary of current status:
- TC17.0 crashes in case of using usb-serial on 1core x86 cpu with vmlinuz + usb-serial-6.18.2-tinycore.tcz from published download location
- TC17.0 seems stable in case of using usb-serial on 1core x86 cpu with self compiled vmlinuz with integrated usb-serial
- self compiled vmlinuz is 200k bigger than vmlinuz from download location, even executing "very clean to-the-book" creation sequence

Additional notes:
I was testing self compiled vmlinuz with published usb-serial-6.18.2-tinycore.tcz to identify whether crashing is cause by extension or by base kernel but that test will need to be repeated because I did need to stop it before being conclusive

As it looks likes the self compiled kernel is more stable than the published kernel AND it has a 200k bigger filesize (so its different for sure), it would be nice to understand where the difference comes from. I can only do that with help/direction from the TC maintainers, @juanito, @paul_123, and/or others I don't know. I will cooperate and follow direction to best of ability if you give that. Given that TC17.1 is in active development I would expect it most handy to explore on TC17.1 rather than TC17.0
==> a first check would probably be to compare compiled tree
If not interested to explore I'm fine as well. I will just use self complied kernel which seems to work fine (it still needs a test over a month to be 100% sure)

So....
Thanks @Rich, @juanito, @Paul_123, @Mocore for helping
This thread is not completely closed but will likely go on a bit lower intensity.
At least no updates until Sunday...monday.

[gadget42]

@Stefann, thanks for taking the time to relate your experiences.

@all, i also would enjoy learning why the 200k size difference occurs and if it is possible to exactly duplicate the original(both size and checksum? values).

wonderful forum! wonderful participants!

keep us posted @Stefann.

[Paul_123]

Rich already pointed out the difference.  He ran "make mrproper" after copying the TC kernel config, which erased the .config file.   So in other words he built a kernel with default options as decided by the kernel team.

Might not be anything wrong with that, other than its bloated.

[Rich]

Hi Paul_123
Actually, he realized what happened, and copied the config a second time:

... What i did:   
 

Code: [Select]

----- Snip -----
tc@hp510:/krubo/work/TC/linux-6.18.28$ cp ../config-6.18.28-tinycore ./.config
tc@hp510:/krubo/work/TC/linux-6.18.28$ make mrproper
  CLEAN   .config
tc@hp510:/krubo/work/TC/linux-6.18.28$ cp ../config-6.18.28-tinycore ./.config
 ----- Snip ----- ...

[Stefann]

Hi Paul_123
Actually, he realized what happened, and copied the config a second time:

In reality..
I was not sure whether mrproper would need the published config so that’s why I loaded it before. (I’m pretty sure this was not needed but I wanted absolute clarity).
I did indeed reload the published config after that as well because that’s what I had read in all advices.
Note…
The “make oldconfig” came back with “no changes”.

So no… I don’t know where the 200k difference comes from.
I did my best to do everything super clean, no shortcuts.

[Paul_123]

Ok, I missed that.  So back to my original statement.  I wonder if there is a problem with your compression setup.

What happens if you try to recompress the kernel  with advdef -z4   

[Stefann]

Ok, I missed that.  So back to my original statement.  I wonder if there is a problem with your compression setup.

What happens if you try to recompress the kernel  with advdef -z4

Ok.. I tried but “advdef not found”.

Please understand it’s a bit tricky as I’m on a trip. I’m doing this from my iPad over ssh (it’s open but I use a passphrase, no password login) That works but it’s only 1 window and typing with onscreen keyboard is not ideal. And the “peek sheets” I normally have are far away.

I guess I need some tcz but don’t know which and would need clear instruction on how to load from terminal as I normally do that with gui.

[Rich]

Hi Stefann
Install  advcomp.tcz.