pc386 BSP/master branch build failure with smp enabled.
karel.gardas at centrum.cz
Sat Sep 12 07:28:55 UTC 2020
On 9/12/20 6:59 AM, Gedare Bloom wrote:
> On Fri, Sep 11, 2020 at 11:53 AM Karel Gardas <karel.gardas at centrum.cz> wrote:
>> On 9/11/20 6:16 PM, Gedare Bloom wrote:
>>> Looks to be coming through
>>> cpukit/score/cpu/i386/include/rtems/asm.h:157 movb
>>> imps_apic_cpu_map(\REG),\REG /* CPU ID in REG */
>>> The assembler is right to complain. movb has to target one of the
>>> 1-byte mnemonics, so this should be %al for the LSB of %eax. One needs
>>> to check through this logic carefully, but I think the right thing to
>>> do here is:
>>> movzbl imps_apic_cpu_map(\REG),\REG
>>> That should do the trick. Can you test it locally?
>> Sure! Builds fine and testsuite testing reveals:
> Great. I opened #4076 to apply this fix on 5 branch.
Wait a minute. I've thought 5 branch is perfectly fine. What's broken is
master. Well, 5.1rc2 and 5.1 release were fine for SMP on pc386 so I
expected 5 branch be too. -- Anyway, I'll test that for sure.
> It should also go
> to master.
That's what IMHO only broken, but will test 5 and report back.
More information about the devel