Problems building "2.4.0-test9" kernel source on/for IA-64

Hello,

I am posting this just in case it has not been already reported and/or fixed thru the latest patch versions (ie somewhere between the base "-test9" source tree version and the current/latest "2.4.0-test10/pre3"). Trying to build "-test9" full tar-ball on/for an IA-64 SDV, I got the following problems/errors :
_ multiple compilation errors for "arch/ia64/kernel/setup.c" source
starting with line #215 and apparently related to "definitions/declarations" issues when it pointed to a routine ("setup_arch()") body !!
In fact this was due to a bug (typo ??) in "setup_arch()" routine itself, where an "if{}else" statement contained an unnecessary "}" ("} else" at line #160 in the original "setup.c" source) finally shortening the routine's body and causing next routines statement to be parsed as "definitions/declarations".
The following fix/diff corrected the problem :
-------------------------

$ diff setup.c.bfi setup.c.orig
160,165c160
< /*            } else
< This accolade appear to be a "typo" which mainly cause the upper "if"
statemen
t
< and the routine itself to close/end prematurelly, thus causing compilation
< errors for the rest of the routines statements now left outside !!
< B.Faccini ([email protected]) */
<               else
---

$
-------------------------

_ "unsatisfied externals" for "[disable,reenable]_usb()" entry
points, in source file "arch/ia64/kernel/irq_ia64.c". This appear to be caused by a joint setup/configuration of "CONFIG_ITANIUM_A1_SPECIFIC"+"CONFIG_USB". Since "[disable,reenable]_usb()" are no longer part of any "-test9" Kernel source, I was only abble to fix this by not using "CONFIG_ITANIUM_A1_SPECIFIC" setting since this build was intended for an A3 stepping system, I assume it should work ...

I did not have already tested this Kernel build, hope it will run/boot ok, and I will post any further issue.

Bruno.

<<Bruno Faccini (E-mail).vcf>>

Bruno> Trying to build "-test9" full tar-ball on/for an IA-64 SDV, I Bruno> got the following problems/errors

Sounds like you didn't apply the IA-64 patch in the .../ports/ia64/ directory. Could this be the problem?

	--david

Ok, I installed "linux-2.4.0-test9-ia64-001004.diff.gz", downloaded from "ftp://ftp.kernel.org/pub/linux/kernel/ports/ia64" and as expected, due to its content, it fixed my earlier 2 issues (multiple compilation errors for "arch/ia64/kernel/setup.c" and "unsatisfied externals" for "[disable,reenable]_usb()" entry points) I got with the original and full "linux-2.4.0-test9" tar-ball.

But the patch/diff file reported a new pb/inconsistencies within the "include/asm-ia64/offsets.h" source file :

--------
patching file linux-2.4.0-test9/include/asm-ia64/offsets.h Hunk #1 FAILED at 11.
Hunk #2 FAILED at 21.
2 out of 2 hunks FAILED -- saving rejects to file linux-2.4.0-test9/include/asm-ia64/offsets.h.rej
--------

and I am a little bit puzzled there due to the suspected high impact ([Task, Thread]'s fields offsets !!) of the related values/defines (expected and new values vs current one) :

------------
- #define IA64_TASK_SIZE 2864 /* 0xb30 */
- #define IA64_TASK_THREAD_OFFSET 896 /* 0x380 */
- #define IA64_TASK_THREAD_KSP_OFFSET 896 /* 0x380 */
- #define IA64_TASK_THREAD_SIGMASK_OFFSET 2744 /* 0xab8 */
+ #define IA64_TASK_SIZE 3328 /* 0xd00 */
+ #define IA64_TASK_THREAD_OFFSET 1424 /* 0x590 */
+ #define IA64_TASK_THREAD_KSP_OFFSET 1424 /* 0x590 */
+ #define IA64_TASK_THREAD_SIGMASK_OFFSET 3184 /* 0xc70 */

vs/with the current source lines !! :

#define IA64_TASK_SIZE 3456 /* 0xd80 */
#define IA64_TASK_THREAD_OFFSET 1456 /* 0x5b0 */
#define IA64_TASK_THREAD_KSP_OFFSET 1456 /* 0x5b0 */
#define IA64_TASK_THREAD_SIGMASK_OFFSET 3304 /* 0xce8 */
-------------

Since this "include/asm-ia64/offsets.h" file appear automatically generated as per the "make config/make dep/make" steps ("arch/ia64/tools/print_offsets") and according to the current configuration options, its actual content may have been computed as part of my earlier build attempt with the "-test9" base distrib., and thus I may ignore the current patch reported errors.

May be I am doing something wrong, but this may also point to a current patching method lack, where config. related and generated files are included within patches/diff files and thus prevent a safe and full install on top of an already used Kernel build source tree.

BTW, and keeping all of this in mind, I re-ran the full Kernel clean-up and build steps and will see how it goes (and mainly if I get a "consistent" and coherent "arch/ia64/tools/print_offsets" file generated as part of the build process) and let you know asap.

Bruno.

Creating a new and "virgin" Kernel source tree with the base "linux-2.4.0-test9" tar-ball and apply "linux-2.4.0-test9-ia64-001004.diff.gz" on top of it worked (ie no "patch" pbs/errors at all) and finally allowed me to build a bootable + reliable Kernel.

But this again looks to me as a possible improvement required with the Kernel distribs+patching process which appear to require an un-touched source tree receiving the base version and any further patches, in parrallel of a second one used for builds and config. changes and to be kept in-sync
....

Bruno.