On 7 March 2013 11:06, Geert Janssens <info@kobaltwit.be> wrote:
Op 18-01-13 22:56, William S Fulton schreef:

On 18/01/13 20:55, Geert Janssens wrote:
On 05-01-13 02:42, William S Fulton wrote:
> On 21/12/12 17:27, Geert Janssens wrote:
>> Hi,
>>
>>
>> So I'm wondering if guile is still officially or unofficially supported
>> by swig ?
>>
>> Note that meanwhile I have added a pointer in the bug tracker item to a
>> patch I propose to fix the guile 2 issues in swig, while maintaining
>> backwards compatibility with both guile 1.6 and 1.8. I'm not sure if
>> this is the preferred way to propose patches. Any feedback would be
>> greatly appreciated.
>>
> I've stopped testing Guile when making releases because it has bit
> rotted and no-one has stepped forward to help out. If you are familiar
> with Guile, which it sounds like you are, then you are welcome to take
> over as maintainer of the language module in as small or large a
> capacity as you like. I don't have time to look at all the language
> modules like Guile, but if whatever changes you make don't regress the
> test-suite for guile 1.6 and 1.8 and improves 2.0, then you can push
> them into the next SWIG release.
>
> William
Well, I'm not fluent in Guile, but I know enough to fix things here and
there.

That's about all that is required.

I am willing to volunteer as a (co-)maintainer for Guile, since it's
currently important for the GnuCash project to get it more up to date.
My time is limited though, so the amount I can contribute can vary.

Well, that's great, I don't think much is going to improve with Guile otherwise.

I have been looking at the test suite of failures for guile. What follows is an impression of my findings so far.

The good news is that the patch I proposed for guile 2 doesn't regress the tests for guile 1.8.

I can't check for guile 1.6  because I don't have access to a guile 1.6 system. And to be honest I don't want to spend my time on 1.6. That version has been abandoned by the guile developers since the release of guile 2.0 so I'd rather focus on the actively supported versions.

The "bad" news is that so far running the test suite with guile 2 gives a lot of errors. It's my impression that guile 2 has a slightly different way of looking for shared libraries. I'll have to investigate further.

Additionally, I note that the guilescm test suite is in much better shape than the guile test suite. The former uses Guile's scm api, while the latter uses the gh api. With guile 1.8 the gh api has been deprecated. In guile 2 it has been completely removed. Again, like support for guile 1.6, I have no intention to revive the gh api test suite. Given its future, I perceive this as a waste of time.

Looking into the details of the test suites for both api's, it turns out that the guilescm test suite is actually heavily dependent on the guile test suite. The difference is only in one config parameter and some tweaks to load scripts from a different directory. Since gh is deprecated, it would make more sense to me that the test suite in guile uses the scm api and that the gh tests are run from a dependent guilegh directory. That way the directories would reflect the preference of api better. It would also be easier to fully drop support for guilegh at some point. Only the guilegh directory would have to be removed.
Are you ok with me rearranging this directory structure ? Obviously I will have to take care not to regress the suites while doing so.


If guile 1.6 is no longer being maintained, then I reckon it is okay to drop support for it. It just needs to be clearly documented.

Feel free to remove the guile-scm test-suite and converge on just one guile test-suite using scm if you think that is best. Although it sounds like you can keep gh supported for the time being. How is the guilegh directory going to work in the test-suite. Will it be setup in the same way that guile and guile-scm are at the moment? I'm very keen that the Makefiles in the test-suite are near identical to all the other target languages for maintainability by those who don't have the first clue about guile.

William