Re: modules/acpi/acpi fails in make universe

From: Ruslan Ermilov (ru_at_FreeBSD.org)
Date: 07/19/05

  • Next message: Ed Schouten: "Re: Errno man page" 
    Date: Tue, 19 Jul 2005 14:22:58 +0300
To: Poul-Henning Kamp <phk@haven.freebsd.dk>

    On Sat, Jul 16, 2005 at 11:16:34AM +0200, Poul-Henning Kamp wrote:
    >
    > When building i386/LINT in "make universe", the acpi/acpi module
    > fails to build:
    >
    > ===> acpi/acpi (depend)
    > Warning: Object directory not changed from original /usr/obj/bang/src0/src/sys/L
    > INT
    > cc -O2 -fno-strict-aliasing -pipe -I. -I@ -c /bang/src0/src/sys/i386/acpica/ac
    > pi_wakecode.S
    > /bang/src0/src/sys/i386/acpica/acpi_wakecode.S:35:19: assym.s: No such file or d
    > irectory
    > /bang/src0/src/sys/i386/acpica/acpi_wakecode.S: Assembler messages:
    > /bang/src0/src/sys/i386/acpica/acpi_wakecode.S:103: Error: suffix or operands in
    > valid for `ljmp'
    > *** Error code 1
    > 1 error
    > *** Error code 2
    > @ -> /bang/src0/src/sys
    >
    >
    > Ad far as I can tell, this bit of sys/modules/acpi/acpi/Makefile
    > is the offending stuff, but I can't tell what's wrong:
    >
    > acpi_wakecode.h: acpi_wakecode.S assym.s
    > ${MAKE} -f ${.CURDIR}/../../../${MACHINE_ARCH}/acpica/Makefile \
    > MAKESRCPATH=${.CURDIR}/../../../${MACHINE_ARCH}/acpica
    >
    Can you reproduce this just building a LINT kernel? I cannot, and
    I fail to see how this could be a problem. The failed "cc" command
    is run from ${.CURDIR}/../../../${MACHINE_ARCH}/acpica/Makefile and
    it only runs after assym.s was created. But I have found a problem
    building assym.s manually which needs fixed kmod.mk:

    %%%
    Index: kmod.mk
    ===================================================================
    RCS file: /home/ncvs/src/sys/conf/kmod.mk,v
    retrieving revision 1.192
    diff -u -r1.192 kmod.mk
    --- kmod.mk 22 Apr 2005 17:39:58 -0000 1.192
    +++ kmod.mk 19 Jul 2005 10:55:55 -0000
    @@ -408,7 +408,10 @@
     assym.s: @/kern/genassym.sh
     .endif
             sh @/kern/genassym.sh genassym.o > ${.TARGET}
    -genassym.o: @/${MACHINE_ARCH}/${MACHINE_ARCH}/genassym.c @ machine
    +.if exists(@)
    +genassym.o: @/${MACHINE_ARCH}/${MACHINE_ARCH}/genassym.c
    +.endif
    +genassym.o: @ machine ${SRCS:Mopt_*.h}
             ${CC} -c ${CFLAGS:N-fno-common} \
                 @/${MACHINE_ARCH}/${MACHINE_ARCH}/genassym.c
     .endif
    %%%

    If you can reproduce the breakage, please put the complete
    combined stdout+stderr output from a failed i386 LINT build
    available somewhere for download.

    Cheers, 

    -- 
Ruslan Ermilov
ru@FreeBSD.org
FreeBSD committer

  • Next message: Ed Schouten: "Re: Errno man page"

    Relevant Pages

    • Re: Computing a*b/c without overflow in the preprocessor
      ... I want to devise a C89 preprocessor macro MULDIVwhich, given positive integer constant expressions a b c, produce an integer constant expression evaluable by the preprocessor and equal to a*b/c even though a*b might grossly overflow the range of long. ... #error "MULDIV1fails" ...
      (comp.lang.c)
    • Re: Computing a*b/c without overflow in the preprocessor
      ... I want to devise a C89 preprocessor macro MULDIVwhich, given positive integer constant expressions a b c, produce an integer constant expression evaluable by the preprocessor and equal to a*b/c even though a*b might grossly overflow the range of long. ... #error "MULDIV1fails" ...
      (comp.lang.c)
    • Re: Outlook 2003 Change
      ... ENDIF ... >Sue Mosher, Outlook MVP ... >> In Outlook 2003 it fails on the send with the message ...
      (microsoft.public.outlook.program_vba)
    • Computing a*b/c without overflow in the preprocessor
      ... I came up with something that works well enough for my current application, by combining two techniques: ... #error "MULDIV1fails" ... I could live with a slightly inaccurate result, provided the relative error is always at most 1/10000. ...
      (comp.lang.c)
    • Re: perf counter issue - WARN_ON_ONCE(!list_empty(&tsk->perf_counter_ctx.counter_list));
      ... run with specific parameters to reproduce it? ... I'll try this with some java benchmarks we have, AMQP related, lets see ... fails, with only the call in wait_task_zombiehackbench works and the ... counter_sched_out(counter, cpuctx, ctx); ...
      (Linux-Kernel)