[Qemu-devel] Make test

[Qemu-devel] Make test

[Programmingkid]

I think 'make test' is an abandoned testing option for QEMU so this report might be unimportant. Here is my results for running 'make test':

$ make test
make -C tests/tcg test
  LINK    test_path
Undefined symbols for architecture x86_64:
  "_buffer_is_zero", referenced from:
      _qemu_iovec_is_zero in test_path.o
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[1]: *** [test_path] Error 1
make: *** [test] Error 2

Re: [Qemu-devel] Make test

[Cleber Rosa]

[-- Attachment #1: Type: text/plain, Size: 1067 bytes --]



On 08/24/2017 03:10 PM, Programmingkid wrote:
> I think 'make test' is an abandoned testing option for QEMU so this report might be unimportant. Here is my results for running 'make test':
> 
> $ make test
> make -C tests/tcg test
>   LINK    test_path
> Undefined symbols for architecture x86_64:
>   "_buffer_is_zero", referenced from:
>       _qemu_iovec_is_zero in test_path.o
> ld: symbol(s) not found for architecture x86_64
> clang: error: linker command failed with exit code 1 (use -v to see invocation)
> make[1]: *** [test_path] Error 1
> make: *** [test] Error 2
> 
> 

I went through the same experience, and reported it (in the form of
patches) here:

http://lists.nongnu.org/archive/html/qemu-devel/2017-08/msg00378.html

Although it looks like you're clang, so it'd be nice to check if these
fixes work for you too.

Regards,

-- 
Cleber Rosa
[ Sr Software Engineer - Virtualization Team - Red Hat ]
[ Avocado Test Framework - avocado-framework.github.io ]
[  7ABB 96EB 8B46 B94D 5E0F  E9BB 657E 8D33 A5F2 09F3  ]


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [Qemu-devel] Make test

[Programmingkid]

> On Aug 24, 2017, at 4:41 PM, Cleber Rosa <crosa@redhat.com> wrote:
> 
> 
> 
> On 08/24/2017 03:10 PM, Programmingkid wrote:
>> I think 'make test' is an abandoned testing option for QEMU so this report might be unimportant. Here is my results for running 'make test':
>> 
>> $ make test
>> make -C tests/tcg test
>>  LINK    test_path
>> Undefined symbols for architecture x86_64:
>>  "_buffer_is_zero", referenced from:
>>      _qemu_iovec_is_zero in test_path.o
>> ld: symbol(s) not found for architecture x86_64
>> clang: error: linker command failed with exit code 1 (use -v to see invocation)
>> make[1]: *** [test_path] Error 1
>> make: *** [test] Error 2
>> 
>> 
> 
> I went through the same experience, and reported it (in the form of
> patches) here:
> 
> http://lists.nongnu.org/archive/html/qemu-devel/2017-08/msg00378.html
> 
> Although it looks like you're clang, so it'd be nice to check if these
> fixes work for you too.
> 
> Regards,
> 
> -- 
> Cleber Rosa
> [ Sr Software Engineer - Virtualization Team - Red Hat ]
> [ Avocado Test Framework - avocado-framework.github.io ]
> [  7ABB 96EB 8B46 B94D 5E0F  E9BB 657E 8D33 A5F2 09F3  ]
> 

Thank you for the patches. I will test them. I was wondering what is the point to having both 'make check' and 'make test'. It looks like everyone is using 'make check'. Maybe we are better off removing the 'make test' target.

Re: [Qemu-devel] Make test

[Cleber Rosa]

[-- Attachment #1: Type: text/plain, Size: 721 bytes --]



On 08/24/2017 05:50 PM, Programmingkid wrote:
> 
> Thank you for the patches. I will test them. I was wondering what is the point to having both 'make check' and 'make test'. It looks like everyone is using 'make check'. Maybe we are better off removing the 'make test' target.
> 

I think the removing the `make test` target is a good thing, if its
tests are either absorbed by an existing `make check*` one.  Now that
depends on what people think the tcg tests (and others) deserve... TLC?
Total annihilation?

-- 
Cleber Rosa
[ Sr Software Engineer - Virtualization Team - Red Hat ]
[ Avocado Test Framework - avocado-framework.github.io ]
[  7ABB 96EB 8B46 B94D 5E0F  E9BB 657E 8D33 A5F2 09F3  ]


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [Qemu-devel] Make test

[Thomas Huth]

[-- Attachment #1: Type: text/plain, Size: 1029 bytes --]

On 25.08.2017 00:16, Cleber Rosa wrote:
> 
> On 08/24/2017 05:50 PM, Programmingkid wrote:
>>
>> Thank you for the patches. I will test them. I was wondering what is the point to having both 'make check' and 'make test'. It looks like everyone is using 'make check'. Maybe we are better off removing the 'make test' target.
>>
> 
> I think the removing the `make test` target is a good thing, if its
> tests are either absorbed by an existing `make check*` one.  Now that
> depends on what people think the tcg tests (and others) deserve... TLC?
> Total annihilation?

Since we do not have very good test coverage for TCG yet, I'd like to
see this rather fixed than removed! Could you please CC: me on future
versions of your patch series (I missed v1)?

And I'd suggest to rename "make test" to "make check-tcg" to align it
with the other check* targets. (I remember the very first time I was
trying to run the QEMU tests and this "check" vs. "test" naming caused
really some confusion for me).

 Thomas


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [Qemu-devel] Make test

[Peter Maydell]

On 25 August 2017 at 06:49, Thomas Huth <thuth@redhat.com> wrote:
> On 25.08.2017 00:16, Cleber Rosa wrote:
>>
>> On 08/24/2017 05:50 PM, Programmingkid wrote:
>>>
>>> Thank you for the patches. I will test them. I was wondering what is the point to having both 'make check' and 'make test'. It looks like everyone is using 'make check'. Maybe we are better off removing the 'make test' target.
>>>
>>
>> I think the removing the `make test` target is a good thing, if its
>> tests are either absorbed by an existing `make check*` one.  Now that
>> depends on what people think the tcg tests (and others) deserve... TLC?
>> Total annihilation?
>
> Since we do not have very good test coverage for TCG yet, I'd like to
> see this rather fixed than removed! Could you please CC: me on future
> versions of your patch series (I missed v1)?

The problem is that we don't have a good framework for building
guest binaries to run under TCG. We should sort out one of those
so that it's easy for a new test to say "this is the .c file,
build it with an ARM gcc [eg using our docker stuff], then
run it with this QEMU command". When we have that it should
be straightforward to convert the 'make test' tests to use
that, and they can be run under 'make check'.

thanks
-- PMM

Re: [Qemu-devel] Make test

[Fam Zheng]

On Fri, 08/25 10:10, Peter Maydell wrote:
> On 25 August 2017 at 06:49, Thomas Huth <thuth@redhat.com> wrote:
> > On 25.08.2017 00:16, Cleber Rosa wrote:
> >>
> >> On 08/24/2017 05:50 PM, Programmingkid wrote:
> >>>
> >>> Thank you for the patches. I will test them. I was wondering what is the point to having both 'make check' and 'make test'. It looks like everyone is using 'make check'. Maybe we are better off removing the 'make test' target.
> >>>
> >>
> >> I think the removing the `make test` target is a good thing, if its
> >> tests are either absorbed by an existing `make check*` one.  Now that
> >> depends on what people think the tcg tests (and others) deserve... TLC?
> >> Total annihilation?
> >
> > Since we do not have very good test coverage for TCG yet, I'd like to
> > see this rather fixed than removed! Could you please CC: me on future
> > versions of your patch series (I missed v1)?
> 
> The problem is that we don't have a good framework for building
> guest binaries to run under TCG. We should sort out one of those
> so that it's easy for a new test to say "this is the .c file,
> build it with an ARM gcc [eg using our docker stuff], then
> run it with this QEMU command". When we have that it should
> be straightforward to convert the 'make test' tests to use
> that, and they can be run under 'make check'.

This seems straigtforward to do with docker. Could you elaborate a concrete
example of one of the "this is the .c file, ..." test so I can try translate it
to a Makefile rule using our docker images?

fam

Re: [Qemu-devel] Make test

[Peter Maydell]

On 25 August 2017 at 10:33, Fam Zheng <famz@redhat.com> wrote:
> On Fri, 08/25 10:10, Peter Maydell wrote:
>> The problem is that we don't have a good framework for building
>> guest binaries to run under TCG. We should sort out one of those
>> so that it's easy for a new test to say "this is the .c file,
>> build it with an ARM gcc [eg using our docker stuff], then
>> run it with this QEMU command". When we have that it should
>> be straightforward to convert the 'make test' tests to use
>> that, and they can be run under 'make check'.
>
> This seems straigtforward to do with docker. Could you elaborate a concrete
> example of one of the "this is the .c file, ..." test so I can try translate it
> to a Makefile rule using our docker images?

tests/tcg/hello-i386.c is probably a good place to start.
It wants to be built with an i386 compiler like so:
i386-linux-gcc -nostdlib -Wall -O2 -g -fno-strict-aliasing -static  -o
hello-i386 hello-i386.c
and then run with
qemu-i386 ./hello-i386
which should print "Hello World".

There's also a hello-arm which you can build with
arm-linux-gnueabihf-gcc -Wall -g -O2 -marm -c -o hello-arm.o hello-arm.c
arm-linux-gnueabihf-ld -o hello-arm hello-arm.o
and run with
qemu-arm ./hello-arm
to print "Hello World". (You can probably compile and link in one
step, but for some reason tests/tcg/Makefile is doing it with
two steps.)

thanks
-- PMM

Re: [Qemu-devel] Make test

[Programmingkid]

> On Aug 24, 2017, at 4:41 PM, Cleber Rosa <crosa@redhat.com> wrote:
> 
> 
> 
> On 08/24/2017 03:10 PM, Programmingkid wrote:
>> I think 'make test' is an abandoned testing option for QEMU so this report might be unimportant. Here is my results for running 'make test':
>> 
>> $ make test
>> make -C tests/tcg test
>>  LINK    test_path
>> Undefined symbols for architecture x86_64:
>>  "_buffer_is_zero", referenced from:
>>      _qemu_iovec_is_zero in test_path.o
>> ld: symbol(s) not found for architecture x86_64
>> clang: error: linker command failed with exit code 1 (use -v to see invocation)
>> make[1]: *** [test_path] Error 1
>> make: *** [test] Error 2
>> 
>> 
> 
> I went through the same experience, and reported it (in the form of
> patches) here:
> 
> http://lists.nongnu.org/archive/html/qemu-devel/2017-08/msg00378.html
> 
> Although it looks like you're clang, so it'd be nice to check if these
> fixes work for you too.
> 
> Regards,
> 
> -- 
> Cleber Rosa
> [ Sr Software Engineer - Virtualization Team - Red Hat ]
> [ Avocado Test Framework - avocado-framework.github.io ]
> [  7ABB 96EB 8B46 B94D 5E0F  E9BB 657E 8D33 A5F2 09F3  ]
> 

After applying the patches I saw this error:

$ make test
make -C tests/tcg test
  CC      test_path.o
  LINK    test_path
./test_path
cc -m32 -nostdlib -Wall -O2 -g -fno-strict-aliasing -static  -o hello-i386 hello-i386.c
hello-i386.c:1:10: fatal error: 'asm/unistd.h' file not found
#include <asm/unistd.h>
         ^
1 error generated.
make[1]: *** [hello-i386] Error 1
make: *** [test] Error 2


The path to this header file is /usr/include/unistd.h and /usr/include/sys/unistd.h on my computer. Maybe you could add this to your patches to make them more Mac-friendly:

#ifdef __MACH__ /* Mach kernel - what's used in Mac OS X */
	#include <unistd.h>
#endif

http://pubs.opengroup.org/onlinepubs/7908799/xsh/unistd.h.html
Then again this page states you should be able to do a "#include <unistd.h>" on your posix compatible platform.

Thank you.

Re: [Qemu-devel] Make test

[Cleber Rosa]

[-- Attachment #1: Type: text/plain, Size: 1280 bytes --]



On 08/24/2017 06:18 PM, Programmingkid wrote:
> After applying the patches I saw this error:
> 
> $ make test
> make -C tests/tcg test
>   CC      test_path.o
>   LINK    test_path
> ./test_path
> cc -m32 -nostdlib -Wall -O2 -g -fno-strict-aliasing -static  -o hello-i386 hello-i386.c
> hello-i386.c:1:10: fatal error: 'asm/unistd.h' file not found
> #include <asm/unistd.h>
>          ^
> 1 error generated.
> make[1]: *** [hello-i386] Error 1
> make: *** [test] Error 2
> 

Thanks for testing it.

> 
> The path to this header file is /usr/include/unistd.h and /usr/include/sys/unistd.h on my computer. Maybe you could add this to your patches to make them more Mac-friendly:
> 
> #ifdef __MACH__ /* Mach kernel - what's used in Mac OS X */
> 	#include <unistd.h>
> #endif
> 
> http://pubs.opengroup.org/onlinepubs/7908799/xsh/unistd.h.html
> Then again this page states you should be able to do a "#include <unistd.h>" on your posix compatible platform.
> 
> Thank you.
> 

I'll take a look and attempt a more portable fix on a v2.

Regards!

-- 
Cleber Rosa
[ Sr Software Engineer - Virtualization Team - Red Hat ]
[ Avocado Test Framework - avocado-framework.github.io ]
[  7ABB 96EB 8B46 B94D 5E0F  E9BB 657E 8D33 A5F2 09F3  ]


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [Qemu-devel] Make test

[Programmingkid]

> On Aug 24, 2017, at 6:21 PM, Cleber Rosa <crosa@redhat.com> wrote:
> 
> 
> 
> On 08/24/2017 06:18 PM, Programmingkid wrote:
>> After applying the patches I saw this error:
>> 
>> $ make test
>> make -C tests/tcg test
>>  CC      test_path.o
>>  LINK    test_path
>> ./test_path
>> cc -m32 -nostdlib -Wall -O2 -g -fno-strict-aliasing -static  -o hello-i386 hello-i386.c
>> hello-i386.c:1:10: fatal error: 'asm/unistd.h' file not found
>> #include <asm/unistd.h>
>>         ^
>> 1 error generated.
>> make[1]: *** [hello-i386] Error 1
>> make: *** [test] Error 2
>> 
> 
> Thanks for testing it.
> 
>> 
>> The path to this header file is /usr/include/unistd.h and /usr/include/sys/unistd.h on my computer. Maybe you could add this to your patches to make them more Mac-friendly:
>> 
>> #ifdef __MACH__ /* Mach kernel - what's used in Mac OS X */
>> 	#include <unistd.h>
>> #endif
>> 
>> http://pubs.opengroup.org/onlinepubs/7908799/xsh/unistd.h.html
>> Then again this page states you should be able to do a "#include <unistd.h>" on your posix compatible platform.
>> 
>> Thank you.
>> 
> 
> I'll take a look and attempt a more portable fix on a v2.
> 
> Regards!
> 
> -- 
> Cleber Rosa
> [ Sr Software Engineer - Virtualization Team - Red Hat ]
> [ Avocado Test Framework - avocado-framework.github.io ]
> [  7ABB 96EB 8B46 B94D 5E0F  E9BB 657E 8D33 A5F2 09F3  ]

Forgot to mention that doing a "#include <unistd.h>" does work on Mac OS X. Does this work on your platform?

#include <stdio.h>
#include <unistd.h>

int main(int argc, char *argv[])
{
	printf("version = %ld\n", _POSIX_VERSION);
	return 0;
}