[syslinux] [PATCH] Fix recognition of keeppxe option

[Gene Cumm]

On Tue, Jun 14, 2016 at 12:33 PM, Ady via Syslinux <syslinux at zytor.com> wrote:
>
>> > kernel.c:new_linux_kernel() to load_linux.c:bios_boot_linux() because
>> > there is no convenient way in new_linux_kernel() to control the boot
>> > flags value.
>>
>> This is the part that has me questioning things and trying to recall
>> if any other KERNEL-like directives ever utilize keeppxe.
>>
>
> @Gene,
>
> Not being a developer myself, I don't understand this "other
> KERNEL-like directives" sentence. I do know what "kernel-like

Those directives which acts like KERNEL, specifying a binary to
execute.  BOOT BSS COM32 COMBOOT CONFIG FDIMAGE LINUX LOCALBOOT PXE.

> directives" means, but I do not know which one is the one you are
> referring to here, and why "_other_ KERNEL-like directives" would be a

I'm referring to all of the above collectively.

> concern. I mean, other than the behavior regarding file name
> extensions, shouldn't the KERNEL-like directives relevant in this case
> be equivalent to each other? (FWIW, I say "relevant" because I am
> assuming that the "CONFIG" and the "LOCALBOOT" directives are not so
> relevant in this context; are they?).

KERNEL guesses what the proper loading method is and should generally
be avoided.  A few years back, a distribution had a kernel that ended
in ".0" while using KERNEL resulting in PXE behavior instead of LINUX.

> One case that has been affected by the lack of KEEPPXE support /
> parsing since Syslinux v.5+ is Reactos, which is supposed to be
> bootable by either mboot.c32 or by chain.c32. In the case of chain.c32,

In my opinion, COM32s should be handling "keeppxe" themselves
programatically, directly or via library calls.  COMBOOT has been
removed at this point but it's questionable whether the core or the
COMBOOT binary should handle "keeppxe".  The remaining KERNEL-like
directives should have "keeppxe" handled by the core, if appropriate.

HPA, thoughts?

> the current default values for Reactos are incorrect, so specifying
> adequate values manually in its command is required, in addition to
> fixing the keeppxe matter. I mention Reactos so to provide - for
> someone else, willing to test - one way of testing the patch, comparing
> the results with the behavior of Syslinux v.4.05 while using the same
> configuration file in both tests.

This makes me think that mboot.c32 and chain.c32 should be enhanced to
provide this.

-- 
-Gene