build failure of drivers/scsi/gdth.c

build failure of drivers/scsi/gdth.c

[Or Gerlitz]

I got into that with 2.6.37 after adding some dynamic debug
configs (I can send the compressed .config if needed)

The same/similar problem happens also for the few more places
which use LINUX_VERSION_CODE

  CHK     include/linux/version.h
  CHK     include/generated/utsrelease.h
  CALL    scripts/checksyscalls.sh
  CHK     include/generated/compile.h
  CC [M]  drivers/scsi/gdth.o
drivers/scsi/gdth.c: In function gdth_ioctl:
drivers/scsi/gdth.c:4510: error: expected expression before >> token
drivers/scsi/gdth.c:4511: error: expected expression before >> token
drivers/scsi/gdth.c:4512: error: invalid lvalue in unary
make[2]: *** [drivers/scsi/gdth.o] Error 1
make[1]: *** [drivers/scsi] Error 2
make: *** [drivers] Error 2

Re: build failure of drivers/scsi/gdth.c

[Randy Dunlap]

On Wed, 12 Jan 2011 12:29:19 +0200 Or Gerlitz wrote:

> I got into that with 2.6.37 after adding some dynamic debug
> configs (I can send the compressed .config if needed)
> 
> The same/similar problem happens also for the few more places
> which use LINUX_VERSION_CODE
> 
>   CHK     include/linux/version.h
>   CHK     include/generated/utsrelease.h
>   CALL    scripts/checksyscalls.sh
>   CHK     include/generated/compile.h
>   CC [M]  drivers/scsi/gdth.o
> drivers/scsi/gdth.c: In function gdth_ioctl:
> drivers/scsi/gdth.c:4510: error: expected expression before >> token
> drivers/scsi/gdth.c:4511: error: expected expression before >> token
> drivers/scsi/gdth.c:4512: error: invalid lvalue in unary
> make[2]: *** [drivers/scsi/gdth.o] Error 1
> make[1]: *** [drivers/scsi] Error 2
> make: *** [drivers] Error 2
> --

I've never seen this problem in 2.6.37 AFAIK and I have 2.6.37 gdth.o and gdth.ko
built on my system now.

How can I reproduce this?

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

Re: build failure of drivers/scsi/gdth.c

[Or Gerlitz]

Randy Dunlap <rdunlap@xenotime.net> wrote:
> I've never seen this problem in 2.6.37 AFAIK and I have 2.6.37 gdth.o and gdth.ko
> built on my system now. How can I reproduce this?

I'll send you now my .config

Re: build failure of drivers/scsi/gdth.c

[Randy Dunlap]

On Thu, 13 Jan 2011 11:20:20 +0200 Or Gerlitz wrote:

> Randy Dunlap <rdunlap@xenotime.net> wrote:
> > I've never seen this problem in 2.6.37 AFAIK and I have 2.6.37 gdth.o and gdth.ko
> > built on my system now. How can I reproduce this?
> 
> I'll send you now my .config
> --

It builds OK for me.

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

Re: build failure of drivers/scsi/gdth.c

[Or Gerlitz]

Randy Dunlap wrote:
> It builds OK for me.

okay, maybe something else is broken on my build environment, I'll send
you off list the compressed verbose build log, the problem I'm not spotting
is why LINUX_VERSION_CODE is defined to nothing in my env. I'm hitting this
error on drivers/scsi/gdth.c and some other code that look on that constant,
luckily there aren't many drivers doing that...

Or.

Re: build failure of drivers/scsi/gdth.c

[Randy Dunlap]

On Tue, 18 Jan 2011 09:39:18 +0200 Or Gerlitz wrote:

> Randy Dunlap wrote:
> > It builds OK for me.
> 
> okay, maybe something else is broken on my build environment, I'll send
> you off list the compressed verbose build log, the problem I'm not spotting
> is why LINUX_VERSION_CODE is defined to nothing in my env. I'm hitting this
> error on drivers/scsi/gdth.c and some other code that look on that constant,
> luckily there aren't many drivers doing that...

Thanks for more data/details.  There was something similar to this reported
a few months ago, but nothing has been done about it.

What shell are you using?
Are you using any other build scripts that set an variable named VERSION ?


See https://bugzilla.kernel.org/show_bug.cgi?id=16510

An alternative to the current generated version.h file was also suggested:

echo "#define LINUX_VERSION_CODE KERNEL_VERSION($(VERSION), $(PATCHLEVEL),
$(SUBLEVEL))"


HTH.
---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

Re: build failure of drivers/scsi/gdth.c

[Or Gerlitz]

Randy Dunlap wrote:
> What shell are you using?
> Are you using any other build scripts that set an variable named VERSION ?
> See https://bugzilla.kernel.org/show_bug.cgi?id=16510

Thanks for the bz pointer, this happens as I wanted to change the kernel version
and added something for sublevel instead for extraversion, I can build 
drivers/scsi/gdth.c fine now, 

so with 

> VERSION = 2
> PATCHLEVEL = 6
> SUBLEVEL = 37
> EXTRAVERSION = -dd
> NAME = Flesh-Eating Bats with Fangs

include/linux/version.h has

> #define LINUX_VERSION_CODE 132645
> #define KERNEL_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))

and with 

> VERSION = 2
> PATCHLEVEL = 6
> SUBLEVEL = 37-dd
> EXTRAVERSION =
> NAME = Flesh-Eating Bats with Fangs


include/linux/version.h has

> #define LINUX_VERSION_CODE
> #define KERNEL_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))

and whoever is addressing LINUX_VERSION_CODE fails to get built