Re: [Buildroot] [PATCH 1/2] package/criu: new package

[Marcus Folkesson <marcus.folkesson@gmail.com>]

[-- Attachment #1.1: Type: text/plain, Size: 5823 bytes --]

Hi Julien,

On Tue, Sep 12, 2023 at 11:53:32PM +0200, Julien Olivain wrote:
> Hi Marcus,
> 
> On 12/09/2023 14:53, Marcus Folkesson wrote:
> > Hi Julien,
> > 
> > 
> > On Sun, Sep 10, 2023 at 11:05:19PM +0200, Julien Olivain wrote:
> > > Hi Marcus,
> > > 
> > > On 10/09/2023 21:41, Marcus Folkesson wrote:
> > > > Hi Julien,
> > > >
> > > > On Sat, Sep 09, 2023 at 11:16:27PM +0200, Julien Olivain wrote:
> > > > > Hi Thomas, Marcus,
> > > > >
> > > > > On 09/09/2023 15:55, Thomas Petazzoni wrote:
> > > > > > On Sat, 9 Sep 2023 15:03:13 +0200
> > > > > > Marcus Folkesson <marcus.folkesson@gmail.com> wrote:
> > > > > >
> > > > > > > Thank you for your solid review.
> > > > >
> > > > > [...]
> > > > >
> > > > > > One good thing would be to have a runtime test for criu, in
> > > > > > support/testing. Julien Olivain (in Cc) can probably provide some
> > > > > > guidance here.
> > > > >
> > > > >
> > > > > Marcus, you should be able to reproduce that with the command:
> > > > >
> > > > >     utils/test-pkg -p criu
> > > >
> > > > I've sent out a v2 where the comments from Thomas is addressed. But I am
> > > > not able to reproduce the errors you got with test-pkg.
> > > >
> > > > I'm not sure why it skips my tests. Could you please provide some
> > > > guidance?
> > > 
> > > All test are skipped because of the "depends on
> > > BR2_PACKAGE_HOST_PYTHON3" not being set by default.
> > > 
> > > I'm not sure is this "depends on" is really needed, as I expect it
> > > to be a build dependency. If it's the case, the dependency is only
> > > needed in the .mk file. I let the reviewers comment the v2 patch
> > > on that topic...
> > > 
> > > Anyways, for the moment, this can be managed by manually creating a
> > > config setting this for test-pkg. For example:
> > > 
> > >     cat > criu.config <<EOF
> > >     BR2_PACKAGE_HOST_PYTHON3=y
> > >     BR2_PACKAGE_CRIU=y
> > >     EOF
> > > 
> > > Then, test-pkg will no longer skip the tests. On my side, I now see:
> > > 
> > >     utils/test-pkg -c criu.config -p criu
> > >                     bootlin-armv5-uclibc [1/6]: FAILED
> > >                      bootlin-armv7-glibc [2/6]: FAILED
> > >                    bootlin-armv7m-uclibc [3/6]: SKIPPED
> > >                      bootlin-x86-64-musl [4/6]: OK
> > >                       br-arm-full-static [5/6]: SKIPPED
> > >                             sourcery-arm [6/6]: FAILED
> > >     6 builds, 2 skipped, 3 build failed, 0 legal-info failed, 0
> > > show-info
> > > failed
> > > 
> > > I let you inspect the build log files ~/br-test-pkg/*/logfile for
> > > the details.
> > 
> > 
> > The error seems to be related to host-libzlib rather than criu :-/
> > 
> > 
> > make[1]: Leaving directory '/home/marcus/mnt/encwork/git/br-test-pkg/bootlin-armv5-uclibc/build/host-libzlib-1.3'
> > ***
> > *** ERROR: package host-libzlib installs executables without proper
> > RPATH:
> > ***   /home/marcus/mnt/encwork/git/br-test-pkg/bootlin-armv5-uclibc/host/bin/xmlwf
> > ***   /home/marcus/mnt/encwork/git/br-test-pkg/bootlin-armv5-uclibc/host/bin/pkgconf
> > ***   /home/marcus/mnt/encwork/git/br-test-pkg/bootlin-armv5-uclibc/host/bin/protoc
> > ***   /home/marcus/mnt/encwork/git/br-test-pkg/bootlin-armv5-uclibc/host/bin/protoc-gen-c
> > ***   /home/marcus/mnt/encwork/git/br-test-pkg/bootlin-armv5-uclibc/host/bin/python3.11
> > ***   /home/marcus/mnt/encwork/git/br-test-pkg/bootlin-armv5-uclibc/host/bin/openssl
> > make: *** [package/pkg-generic.mk:305: /home/marcus/mnt/encwork/git/br-test-pkg/bootlin-armv5-uclibc/build/host-libzlib-1.3/.stamp_host_installed]
> > Error 1
> 
> For information, on my side, I see the criu package build failing
> for various reasons.
> 
> For example, toolchain bootlin-armv5-uclibc is failing with:
> 
> compel/arch/arm/src/lib/infect.c: In function ‘compel_set_task_ext_regs’:
> compel/arch/arm/src/lib/infect.c:107:20: error: ‘PTRACE_SETVFPREGS’
> undeclared (first use in this function); did you mean ‘PTRACE_SETFPREGS’?
>   107 |         if (ptrace(PTRACE_SETVFPREGS, pid, NULL, ext_regs)) {
>       |                    ^~~~~~~~~~~~~~~~~
>       |                    PTRACE_SETFPREGS

I've replaced

config BR2_PACKAGE_CRIU_ARCH_SUPPORTS
	bool
	default y if BR2_arm
	...

With:

config BR2_PACKAGE_CRIU_ARCH_SUPPORTS
	bool
	default y if BR2_ARM_CPU_ARMV6
	default y if BR2_ARM_CPU_ARMV7A
	default y if BR2_ARM_CPU_ARMV7M
	default y if BR2_ARM_CPU_ARMV8A
	default y if BR2_ARM_CPU_ARMV8M


Due to this comment from the yocto recipe:

#
# CRIU just can be built on ARMv7 and ARMv6, so the Makefile check
# if the ARCH is ARMv7 or ARMv6.
#

So the toolchain bootlin-armv5-uclibc will be skipped.


> 
> Toolchain bootlin-armv7-glibc failing with:
> 
> /br-test-pkg/bootlin-armv7-glibc/host/bin/arm-linux-ld:
> ./compel/plugins/std.lib.a(log.o): in function `std_vprint_num':
> log.c:(.text+0x2c0): undefined reference to `__aeabi_idivmod'
> /br-test-pkg/bootlin-armv7-glibc/host/bin/arm-linux-ld:
> ./compel/plugins/std.lib.a(log.o): in function `sbuf_printf':
> log.c:(.text+0x41c): undefined reference to `__aeabi_idivmod'
> /br-test-pkg/bootlin-armv7-glibc/host/bin/arm-linux-ld:
> ./compel/plugins/std.lib.a(string.o): in function `std_vdprintf':
> string.c:(.text+0x114): undefined reference to `__aeabi_idivmod'
> 
> There is possibly others.  I'll try to investigate how to resolve those
> issues.
> 
> > > With this v2, I am now able to build criu on some x86_64 and Aarch64
> > > configurations, so I can start writing a runtime test ;)
> > 
> > Yeay!
> 
> I have now a fully working test on Aarch64. I'll send it when criu
> will be merged.
> 
> Best regards,
> 
> Julien.

/Marcus

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[-- Attachment #2: Type: text/plain, Size: 150 bytes --]

_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot