Skip to contentSkip to conversation

kbuild 2.5 preventing mixture of compilers

DISCUSSION Keith Owens 3 participants 3 messages Nov 1, 2001

lore
Share & tools
lore
#0 lore raw

FYI, kbuild 2.5 will check that all the kernel and modules were compiled with the same version of gcc, in particular they must all have the same value of

  $(CC) -v 2>&1 | sed -ne '1s:.*/\([^/]*/[^/]*\)/[^/]\+$:\1:p'

e.g. i386-redhat-linux/2.96

Minor version data such as build date is assumed to be irrelevant. Anybody who makes significant changes to compiler output without changing the version number gets what they deserve.

Modules built with a different compiler from the kernel will not load unless they are forced with insmod -f.

Is this going to cause problems for anybody? I see no justification for mixing kernel objects built by different compilers and I know of problems that have been caused by doing so.

#1 lore raw

IMHO, that is a good idea... almost.

What about companies that maintain closed source driver modules for their hardware?

I know a lot of people here will say, 'well they should open source them then'. However, some companies don't want to for their own reasons and this 'could' blow them out the water a bit and affect end users.

Quoted 1 line

From a stability point of view, this may seem like an excellent idea. I'm

just scared that this could put companies off producing driver modules for their hardware.

Lee Packham

Quoted 24 lines

FYI, kbuild 2.5 will check that all the kernel and modules were compiled with the same version of gcc, in particular they must all have the same value of

  $(CC) -v 2>&1 | sed -ne '1s:.*/\([^/]*/[^/]*\)/[^/]\+$:\1:p'

e.g. i386-redhat-linux/2.96

Minor version data such as build date is assumed to be irrelevant. Anybody who makes significant changes to compiler output without changing the version number gets what they deserve.

Modules built with a different compiler from the kernel will not load unless they are forced with insmod -f.

Is this going to cause problems for anybody? I see no justification for mixing kernel objects built by different compilers and I know of problems that have been caused by doing so.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#2 lore raw
Quoted 8 lines · In article <[email protected]> you wrote:

IMHO, that is a good idea... almost.

What about companies that maintain closed source driver modules for their hardware?

I know a lot of people here will say, 'well they should open source them then'. However, some companies don't want to for their own reasons and this 'could' blow them out the water a bit and affect end users.

If a different compiler version is known to break (and Keith says he has seen that in practice, and I can see it happen as well given that a few kernel headers depend on compiler version), the vendor in question is better off being informed about the incompatibility. He _already_ has to have an exact match on the kernel version and the other symbols, so adding the compiler to that is not an extra burden, just a "oh we goofed" prevention..