Re: [PATCH v3] configure: Let SLOF be initialized by ./scripts/git-submodule.sh

[Thomas Huth <thuth@redhat.com>]

On 15/06/2020 11.19, Daniel P. Berrangé wrote:
> On Mon, Jun 15, 2020 at 09:49:19AM +0200, Philippe Mathieu-Daudé wrote:
>> The git-submodule.sh script is called by make and initialize the
>> submodules listed in the GIT_SUBMODULES variable generated by
>> ./configure.
>>
>> SLOF is required for building the s390-ccw firmware on s390x, since
>> it is using the libnet code from SLOF for network booting.
>>
>> Add it to the GIT_SUBMODULES when buildint the s390-ccw firmware,
>> to fix:
>>
>>   $ ( cd ${SRC_DIR} ; git submodule update --init roms/SLOF )
>>   Submodule 'roms/SLOF' (https://git.qemu.org/git/SLOF.git) registered for path 'roms/SLOF'
>>   Cloning into '/home/travis/build/user/qemu/roms/SLOF'...
>>   fatal: unable to access 'https://git.qemu.org/git/SLOF.git/': Could not resolve host: git.qemu.org
>>   fatal: clone of 'https://git.qemu.org/git/SLOF.git' into submodule path '/home/travis/build/user/qemu/roms/SLOF' failed
>>   Failed to clone 'roms/SLOF'. Retry scheduled
>>   Cloning into '/home/travis/build/user/qemu/roms/SLOF'...
>>   fatal: unable to access 'https://git.qemu.org/git/SLOF.git/': Could not resolve host: git.qemu.org
>>   fatal: clone of 'https://git.qemu.org/git/SLOF.git' into submodule path '/home/travis/build/user/qemu/roms/SLOF' failed
>>   Failed to clone 'roms/SLOF' a second time, aborting
>>   The command "( cd ${SRC_DIR} ; git submodule update --init roms/SLOF )" exited with 1.
> 
> The "git-submodule.sh" script just invokes "git submodule". So if
> you are getting DNS failures from "git submodule", using git-submodule.sh
> instead is not going to fix the problem.
> 
>>
>> Reported-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
>> Suggested-by: Thomas Huth <thuth@redhat.com>
>> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
>> ---
>>  configure   | 5 +++++
>>  .travis.yml | 1 -
>>  2 files changed, 5 insertions(+), 1 deletion(-)
>>
>> diff --git a/configure b/configure
>> index bb7fd12612..927e4a3d06 100755
>> --- a/configure
>> +++ b/configure
>> @@ -6533,6 +6533,11 @@ if test "$cpu" = "s390x" ; then
>>    write_c_skeleton
>>    if compile_prog "-march=z900" ""; then
>>      roms="$roms s390-ccw"
>> +    # SLOF is required for building the s390-ccw firmware on s390x,
>> +    # since it is using the libnet code from SLOF for network booting.
>> +    if test -e "${source_path}/.git" ; then
>> +      git_submodules="${git_submodules} roms/SLOF"
>> +    fi
>>    fi
>>  fi
> 
> This whole bit of configure looks a bit dubious.
> 
> For all the other firmware images we ship as pre-built blobs, we never
> try to re-build them even if the host compiler supports it.

All the other firmwares have separate projects and repositories, so it's
normal that we don't try to rebuild them from the QEMU repo. But the
s390-ccw firmware is part of the QEMU repository (for some good
reasons), so of course we should build it during the QEMU build process
if possible. And no, please don't suggest to move it to a separate
repository instead, IIRC we've had this discussion in the past already,
and it is easier if it stays in the current place.

> So I don't think we need to make SLOF special.

We are not talking about SLOF here. We are talking about the s390-ccw
firmware which is part of the QEMU repository (see pc-bios/s390-ccw). It
just requires the SLOF submodule for building its network booting part,
since it re-uses the network stack from SLOF.

 Thomas