Re: Porting ruby to 2.6.5... partialy succes here either :-)

[Zoltan B. <zb...@fr...>]

Hi,

if you are interested, here's the patches that were needed
to build and run 2.6.6-rc2-mm1-ruby. (ruby-2.6.5.patch omitted.)

Patch against 2.6.6-rc2 so ruby-2.6.5 generates only trivial rejects:
2.6.6-rc2-to-be-reverted-in-ruby.patch

This is mainly about whitespace fixes and cast removal. vt_ioctl.c
has one less empty line at the end in mainline, it causes a 80+ kB
reject.

"2.6.6-rc2-to-be-reverted-in-ruby-1.patch" is the part of that
that needed thinking. E.g. is it needed to move vcs_init() in
vty_init()? Anyway, I applied it and the locking change in
con_close()/vt_close().

The con_open() / vt_open() change is not clear: placing everything
inside of "if (tty->count =3D=3D 1) { ... }" and returning error code
instead of 0 always. May also be needed in ruby. (?)

Then ruby-2.6.5.patch was applied, drivers/pci/proc.c in mainline
already has the same locking fix that is in ruby, only the semaphor
is kept in a variable instead of evaluating the pointers twice.
The Makefile version reject was trivial. :-)

Then 2.6.6-rc2-mm1 was applied, keyboard.c has a trivial reject,
and nothing else! (Harmless and does nothing without KGDB.)

Here is the patch against the ruby-2.6 tree so it is now against
linux-2.6.6-rc2-mm1.

I also attached the patch against kernel/softirq.c. Needed to boot. :-)

Best regards,
Zolt=E1n B=F6sz=F6rm=E9nyi

Zoltan Boszormenyi =EDrta:
> I also thank you. This two-liner made it possible to run
> 2.6.6-rc2-mm1-ruby now. :-)