Paolo, there's a Make puzzle for you below.
Alistair Francis <alistair.fran...@xilinx.com> writes:
> On Mon, Aug 14, 2017 at 6:34 AM, Markus Armbruster <arm...@redhat.com> wrote:
>> PATCH 3/5 has the exact same subject. Why are the two separate?
>
> You are right, that is a mess.
>
> This one doesn't check for newlines at the end while the earlier one
> checked for and removed new lines.
>
>>
>> Alistair Francis <alistair.fran...@xilinx.com> writes:
[...]
>>> diff --git a/tests/Makefile.include b/tests/Makefile.include
>>> index 7af278db55..4886caf565 100644
>>> --- a/tests/Makefile.include
>>> +++ b/tests/Makefile.include
>>> @@ -560,8 +560,8 @@ tests/test-thread-pool$(EXESUF):
>>> tests/test-thread-pool.o $(test-block-obj-y)
>>> tests/test-iov$(EXESUF): tests/test-iov.o $(test-util-obj-y)
>>> tests/test-hbitmap$(EXESUF): tests/test-hbitmap.o $(test-util-obj-y)
>>> $(test-crypto-obj-y)
>>> tests/test-x86-cpuid$(EXESUF): tests/test-x86-cpuid.o
>>> -tests/test-xbzrle$(EXESUF): tests/test-xbzrle.o migration/xbzrle.o
>>> migration/page_cache.o $(test-util-obj-y)
>>> -tests/test-cutils$(EXESUF): tests/test-cutils.o util/cutils.o
>>> +tests/test-xbzrle$(EXESUF): tests/test-xbzrle.o migration/xbzrle.o
>>> migration/page_cache.o $(test-qom-obj-y)
>>> +tests/test-cutils$(EXESUF): tests/test-cutils.o util/cutils.o
>>> $(test-qom-obj-y)
>>
>> No. What symbols exactly is the linker missing?
>
> Without the change, this is the error I see when running make check:
>
> CC tests/test-x86-cpuid.o
> LINK tests/test-x86-cpuid
> GTESTER tests/test-x86-cpuid
> CC tests/test-xbzrle.o
> LINK tests/test-xbzrle
> libqemustub.a(monitor.o): In function `monitor_get_fd':
> /scratch/alistai/master-qemu/stubs/monitor.c:10: undefined reference
> to `error_setg_internal'
> collect2: error: ld returned 1 exit status
> /scratch/alistai/master-qemu/rules.mak:121: recipe for target
> 'tests/test-xbzrle' failed
> make: *** [tests/test-xbzrle] Error 1
The root cause is "obvious" ;-P
The linker searches each .a just once. We link with
test-util-obj-y = libqemuutil.a libqemustub.a
i.e. the linker first pulls whatever it needs out of libqemuutil.a, then
moves on to pull whatever it now needs out of libqemustub.a. If
libqemustub.a needs something from libqemuutil.a that hasn't been pulled
already, linking fails.
The linker explains its doings (--print-map):
Archive member included to satisfy reference by file (symbol)
libqemuutil.a(cutils.o) tests/test-xbzrle.o (uleb128_encode_small)
libqemuutil.a(qemu-error.o) libqemuutil.a(cutils.o) (warn_report)
libqemustub.a(error-printf.o)
libqemuutil.a(qemu-error.o) (error_vprintf)
libqemustub.a(monitor.o) libqemuutil.a(qemu-error.o) (cur_mon)
Let me explain the linker's explanation:
test-xbzrle.o pulls in libqemuutil.a(cutils.o) for
uleb128_encode_small
libqemuutil.a(cutils.o) pulls in libqemuutil.a(qemu-error.o) for
warn_report
libqemuutil.a(qemu-error.o) pulls in libqemustub.a(error_printf.o)
and libqemustub.a(monitor.o) for error_vprintf and cur_mon
libqemuutil(monitor.o) references error_setg_internal, which can't
be resolved.
The stupid fix is to repeat libraries until the link succeeds:
test-util-obj-y = libqemuutil.a libqemustub.a libqemuutil.a
You may have seen this with -lX11 if you're old enough.
ld(1) suggests the linker can do it for us:
-( archives -)
--start-group archives --end-group
The archives should be a list of archive files. They may be either
explicit file names, or -l options.
The specified archives are searched repeatedly until no new
undefined references are created. Normally, an archive is searched
only once in the order that it is specified on the command line.
If a symbol in that archive is needed to resolve an undefined
symbol referred to by an object in an archive that appears later on
the command line, the linker would not be able to resolve that
reference. By grouping the archives, they all be searched
repeatedly until all possible references are resolved.
Using this option has a significant performance cost. It is best
to use it only when there are unavoidable circular references
between two or more archives.
Sticking '-Wp,-(' and '-Wp,-)' into the command line I get from make V=1
doesn't work for me, though.
The smart solution is not to have .a reference each other.
Paolo, what do you think?
> If only the xbzrle change is made then I see this:
>
> LINK tests/test-xbzrle
> GTESTER tests/test-xbzrle
> CC tests/test-vmstate.o
> LINK tests/test-vmstate
> GTESTER tests/test-vmstate
> CC tests/test-cutils.o
> LINK tests/test-cutils
> util/cutils.o: In function `parse_debug_env':
> /scratch/alistai/master-qemu/util/cutils.c:605: undefined reference to
> `warn_report'
> collect2: error: ld returned 1 exit status
> /scratch/alistai/master-qemu/rules.mak:121: recipe for target
> 'tests/test-cutils' failed
> make: *** [tests/test-cutils] Error 1
