robotpkg

Thanks for the push, and the fixes !

I had moved to `make package` in my test scripts, but I mistakenly reverted that :(

My rebuild of the same fixes is not done yet, but I guess we can go for those low hanging fruit already.

Also, I forgot to update you on the push, because I was dealing with
the redmine app being in a weird state (hence the missing commit
notification after the first push of your commits).

Anyway, no big deal, that was trivial issues to fix, this time :)

Does it make sense to bump

HPP_MIN_VERSION=        6

This would fix this breakage, which is due to a mix of available
versions (hpp-utils-6.0.0 is OK on Rocky8, but hpp-pinocchio is stuck
at 5.0.0 due to dependency requirements changes):
py36-qt5-hpp-practicals-6.0.0

We could do that, yes, but we need to take care of gepetto-viewer which is following HPP versions but is still in v5.
I can release it in v6 and avoid that issue, but not before tomorrow

You know better than I the implications. No urgency anyway.

But the alternative is to figure out why cmake decides that version 6.0.0
is not compatible with version 5.0.0:

CMake Error at /usr/share/cmake/Modules/CMakeFindDependencyMacro.cmake:76 (find_package):
Could not find a configuration file for package "hpp-util" that is
compatible with requested version "5.0.0".

The following configuration files were considered but not accepted:

/opt/openrobots/lib/cmake/hpp-util/hpp-utilConfig.cmake, version: 6.0.0
Call Stack (most recent call first):
/opt/openrobots/lib/cmake/hpp-pinocchio/hpp-pinocchioConfig.cmake:151 (find_dependency)

First, I see no "requested version" in the hpp-pinocchio cmake Config
file that just does find_dependency(hpp-util REQUIRED) and second I can't
find anything saying something about versions compatibility.

I'm still curious to understand what's going on there.

Ah, duck debugging, I found that it is actually hpp-util itself doing this
in hpp-utilConfig.cmake

set(PACKAGE_VERSION "5.0.0")
[...]
if(PACKAGE_FIND_VERSION_MAJOR STREQUAL CVF_VERSION_MAJOR)
set(PACKAGE_VERSION_COMPATIBLE TRUE)
else()
set(PACKAGE_VERSION_COMPATIBLE FALSE)
endif()

If this constraint is actually true, this should be reflected in the
depend.mk of hpp-util.
But should not this rather be a 'version_less' test?

  if(PACKAGE_FIND_VERSION_MAJOR VERSION_LESS CVF_VERSION_MAJOR)
set(PACKAGE_VERSION_COMPATIBLE TRUE)
else()
set(PACKAGE_VERSION_COMPATIBLE FALSE)
endif()

Because in the case no package version is requested in find_package,
cmake sets PACKAGE_FIND_VERSION_MAJOR to 0, so the strequal test will
never match.

I don't understand where this version is coming from. I have:

$ rg 'PACKAGE_VERSION '
lib/cmake/hpp-util/hpp-utilConfigVersion.cmake
12:set(PACKAGE_VERSION "6.0.0")

Oh, yes, "due to dependency requirements changes". Hum, I think in this particular case we could patch to relax the constraint.

Hello,

I did add some commits to finish with HPP v6 in gepetto-viewer, gepetto-viewer-corba, HPP_MIN_VERSION ; and also I'm adding hpp-python, a new project boost-python which is there to allow us to finally drop CORBA / omniORB things in HPP future.
There is also a minor pinocchio update.

Hello !

Did you had a look at those 5 additional commits ?

Thanks :)

It seems that I missed this notification :/

I'll push that today, sorry for the delay.