ubuntu arm

[johnnynunez]

This pull request includes updates to the GitHub Actions workflow configuration to add support for Ubuntu ARM architectures. The most important changes include adding new job configurations for Ubuntu ARM and setting up CastXML for these new environments.

Updates to GitHub Actions workflow:

• .github/workflows/tests.yml: Added job configurations for ubuntu-22.04-arm and ubuntu-24.04-arm with specific settings for architecture, compiler, and Python version.
• .github/workflows/tests.yml: Added steps to set up CastXML for ubuntu-22.04-arm and ubuntu-24.04-arm environments, including downloading and extracting the appropriate binaries.

[johnnynunez]

new bugs appear @iMichka

[iMichka]

Ok that's 7 errors, some of them I think I already know how to fix (it's mostly badly written test that I wrote a long time ago ...). It will take a few days to go through these errors, I'll open separate pull requests to fix them.

[johnnynunez]

Ok that's 7 errors, some of them I think I already know how to fix (it's mostly badly written test that I wrote a long time ago ...). It will take a few days to go through these errors, I'll open separate pull requests to fix them.

Super!

[iMichka]

Fixed some tests with #250, this can now be rebased

[iMichka]

I added a fix here to see if this helps. I think there is just one more declaration to remove.

[johnnynunez]

I added a fix here to see if this helps. I think there is just one more declaration to remove.

Super. Only know 2 failed

[iMichka]

Ok, I have extracted that fix to a separate PR to get it shipped ASAP: #251

[johnnynunez]

Seems that is not installed. I will check

[iMichka]

Yes for one of them castxml is not installed.
I'll push a few more commits here to try to debug the other pygccxml bug

[iMichka]

It's odd, <class 'pygccxml.declarations.namespace.namespace_t'> std [namespace] std has been found twice.
I'll probably have to check the xml file generated by castxml. This will need some more time. I'll continue tomorrow.

If you want to push some more fixes for the missing castxml in the third build, go ahead :)

[johnnynunez]

there was a typo error @iMichka fixed. It should build for 3 arm os

[johnnynunez]

It's odd, <class 'pygccxml.declarations.namespace.namespace_t'> std [namespace] std has been found twice. I'll probably have to check the xml file generated by castxml. This will need some more time. I'll continue tomorrow.

If you want to push some more fixes for the missing castxml in the third build, go ahead :)

Done!

[iMichka]

Here is the xml file based on the CI output https://gist.github.com/iMichka/1b35b9b8c3ac40c59b8e0c0094715971

<Namespace id="_3536" name="std" context="_1"/>
<Namespace id="_66" name="std" context="_1" ....

_3536 is related to __va_list, so that's intriguing

 <Struct id="_1102" name="__va_list" context="_3536" location="f0:0" file="f0" 

[johnnynunez]

I've try everything in my fork, and I didn't find it where is the problem @iMichka

I investigate it, and file by file for me is working with this code:

@staticmethod
    def _join_top_namespaces(main_ns_list, other_ns_list):
        answer = main_ns_list[:]
        for other_ns in other_ns_list:
            # Instead of find_declaration(...), gather ALL that match name
            same_name_namespaces = pygccxml.declarations.find_all_declarations(
                answer,
                decl_type=pygccxml.declarations.namespace_t,
                name=other_ns.name,
                parent=None,  # top-level only
                recursive=False
            )
            if len(same_name_namespaces) == 0:
                answer.append(other_ns)
            elif len(same_name_namespaces) == 1:
                same_name_namespaces[0].take_parenting(other_ns)
            else:
                primary_ns = same_name_namespaces[0]
                for extra_ns in same_name_namespaces[1:]:
                    primary_ns.take_parenting(extra_ns)
                    answer.remove(extra_ns)
                # then unify the new other_ns
                primary_ns.take_parenting(other_ns)

        return answer

I've seen that all tests are working:
"global_ns_fixture_file_by_file1",
"global_ns_fixture_file_by_file2",

the problem are:
"global_ns_fixture_all_at_once1",
"global_ns_fixture_all_at_once2",

[iMichka]

Found it, the same change was missing from __parse_all_at_once.

Looks like the join logic was never really tested / we did not expect this edge case.

This pull request will need a big cleanup. I'll have some time tomorrow evening to move this forward. We are close to be done here :)

Regarding the new implementation of _join_top_namespaces. I see what you did here, it "works" but I feel this can refactored even more. The logic is ... complex to understand. There is a loop with multiple conditions, using append/remove on a list.
I am wondering if now that we use pygccxml.declarations.find_all_declarations, we could not simplify this even more.

[johnnynunez]

Found it, the same change was missing from __parse_all_at_once.

Looks like the join logic was never really tested / we did not expect this edge case.

This pull request will need a big cleanup. I'll have some time tomorrow evening to move this forward. We are close to be done here :)

Regarding the new implementation of _join_top_namespaces. I see what you did here, it "works" but I feel this can refactored even more. The logic is ... complex to understand. There is a loop with multiple conditions, using append/remove on a list. I am wondering if now that we use pygccxml.declarations.find_all_declarations, we could not simplify this even more.

Thanks a lot!

[johnnynunez]

Found it, the same change was missing from __parse_all_at_once.

Looks like the join logic was never really tested / we did not expect this edge case.

This pull request will need a big cleanup. I'll have some time tomorrow evening to move this forward. We are close to be done here :)

Regarding the new implementation of _join_top_namespaces. I see what you did here, it "works" but I feel this can refactored even more. The logic is ... complex to understand. There is a loop with multiple conditions, using append/remove on a list. I am wondering if now that we use pygccxml.declarations.find_all_declarations, we could not simplify this even more.

I can clean it with your fixes.

[johnnynunez]

@iMichka cleaned

[johnnynunez]

@iMichka could you merge this branch first, and then create optimizations in other branches?
Because with this branch, now libraries like OMPL that there was that bug from years, now it is possible to build thanks to your work