(return-from south-africa 'married)

[sross]

Hi all,
I have just returned to London a Married Man and will be back and
working on mudballs shortly, I hope that nothing untoward happened in
my absence.



Out of curiosity, with a show of hands, how many people are currently
using Mudballs on a daily basis?

If you are, what are the issues you are running into? What annoys you
and what behaviour makes you think I'm just not paying attention?

If not, is there a particular reason? I'm really trying to find the
biggest stumbling blocks when it comes to adopting Mudballs.


Cheers,
Sean.

[gugamilare]

Hi, Sean,

Congratulations on your marriage.

Well, I'm new here, and I just got mudballs right now (less than one
hour ago). I currently use extensively clbuild (which is a very good
script that automatically fetches and updates repositories (sometimes
tarballs - including their dependencies) automatically). And one
functionality that I am missing in mudballs is the ability to easily
include a new system. In clbuild, you open my-projects file and write
a new line containing the information about how to fetch the
repository or tarball, like this:

defclass-star get_darcs http://common-lisp.net/project/defclass-star/darcs/defclass-star

I believe this is not possible yet with mudballs, is it? Even if I
need manually to write a callback function or method or something that
fetches and updates the repository.

Cheers.
Gustavo.

[Tobias Rautenkranz]

Hi Sean,

Glad you made it back ;)

I'm currently using mudballs for cl-smoke:
http://tobias.rautenkranz.ch/lisp/cl-smoke/

Mudballs has all the packages I need; response to bug reports is good. There
are however some minor things I have run into:

- mb:document has problems with none alphanumeric characters in symbols. The
functions CXX:* CXX:/ etc. all link to the same file name.
http://tobias.rautenkranz.ch/lisp/cl-smoke/api-doc/cxx/cxx-package.html [776K]

- The generic function foo has the file name "foo-generic-function.html" but if
it were an ordinary function it would be "foo-function.html".
I'd like it to have only one ending for all symbols in the function namespace
and an other one for variables.(So I don't have to think whether it's a
generic function or not.)
And this will probably clash:
(defun foo-generic ())
(defgeneric foo ())

- To allow multiple files with dependencies for mb:test I made a separate test
system (qt.tests). And I actually like it that way.

- There is no simple way to have mudballs search the current directory first?
Solved this by loading the .mbd file with LOAD.
(see: http://tobias.rautenkranz.ch/lisp/cl-smoke/qt.tests/test.lisp)
I need this to run mb:test on darcs record.

- My sysdef.cmake needs a "forward declaration" for the file classes it defines
in the systems using it.
(Example in http://tobias.rautenkranz.ch/lisp/cl-smoke/qt/qt.mbd)
The problem is that sysdef.cmake exports a cmake-file class to be used as a file
class in define-system. But when that define-system is loaded on mudballs
startup the cmake-file class is not defined. Thus one needs to defclass
cmake-file before the define-system.

Is sysdef.cmake a good name? I will contact you for adding cl-smoke to the
mudballs distribution when it's in a releasable state.

[Paul Onions]

On Apr 16, 7:48 pm, sross <ros...@gmail.com> wrote:
> Out of curiosity, with a show of hands, how many people are currently
> using Mudballs on a daily basis?

I use Mudballs in a small way at work and on a personal project I'm
working on.

I'm really only a dabbler in Lisp at the moment (time doesn't permit
much more than that), but the only reason I've found to step outside
the Mudballs environment is to install CLOCC, and that was just to get
access to the CLLIB/math routines (specifically an implementation of
the ERF function). It would be nice if I could get this in Mudballs,
but:
1. CLOCC has its own build system and porting this to Mudballs looks
decidedly non-trivial to me.
2. The only other place I've found an implementation of ERF is in
CLMath, but the licence seems to be for non-commercial use only -- not
good for me.

I would be willing to port a simple (open) math library to Mudballs
for others to use, but I don't think I'm expert enough at Lisp yet to
attempt this with CLOCC.

Paul

[Sean Ross]

It's not currently possible but this is something that I am
considering for future versions of mudballs.
Thanks for you input.

sean.

[Leslie P. Polzer]

On Apr 16, 8:48 pm, sross <ros...@gmail.com> wrote:

>  Out of curiosity, with a show of hands, how many people are currently
> using Mudballs on a daily basis?
>
>  If you are, what are the issues you are running into? What annoys you
> and what behaviour makes you think I'm just not paying attention?

I'm not a regular user right now but I'm soon going to try to use
Mudballs
as a CL package manager for Arch Linux, so you might get some
complaints
then from me. :)

[Sean Ross]

On 17 Apr 2009, at 17:20, Tobias Rautenkranz wrote:

> Hi Sean,
>
> Glad you made it back ;)
>
> I'm currently using mudballs for cl-smoke:
> http://tobias.rautenkranz.ch/lisp/cl-smoke/
>
> Mudballs has all the packages I need; response to bug reports is
> good. There
> are however some minor things I have run into:
>
> - mb:document has problems with none alphanumeric characters in
> symbols. The
> functions CXX:* CXX:/ etc. all link to the same file name.
> http://tobias.rautenkranz.ch/lisp/cl-smoke/api-doc/cxx/cxx-package.html
> [776K]
>

Thanks, I'll raise a bug for this.

> - The generic function foo has the file name "foo-generic-
> function.html" but if
> it were an ordinary function it would be "foo-function.html".
> I'd like it to have only one ending for all symbols in the function
> namespace
> and an other one for variables.(So I don't have to think whether
> it's a
> generic function or not.)
> And this will probably clash:
> (defun foo-generic ())
> (defgeneric foo ())

I'll get this patched as soon as I can, thanks for catching this.

>
> - To allow multiple files with dependencies for mb:test I made a
> separate test
> system (qt.tests). And I actually like it that way.
>

Yes, the current setup for tests, while simple, doesn't allow any
flexibility for users.
I'm entirely convinced that adding a new system for the tests is the
best way to go
as it has a tendency of polluting the searchable system namespace
(although I'm
currently at a loss for a better approach)

> - There is no simple way to have mudballs search the current
> directory first?
> Solved this by loading the .mbd file with LOAD.
> (see: http://tobias.rautenkranz.ch/lisp/cl-smoke/qt.tests/test.lisp)
> I need this to run mb:test on darcs record.
>

Let me give this some thought as I'm not quite sure what the correct
approach is here.
My understanding is that you want to load a system which, rather than
being identified by
version, is identified by location, is this correct?

> - My sysdef.cmake needs a "forward declaration" for the file classes
> it defines
> in the systems using it.
> (Example in http://tobias.rautenkranz.ch/lisp/cl-smoke/qt/qt.mbd)
> The problem is that sysdef.cmake exports a cmake-file class to be
> used as a file
> class in define-system. But when that define-system is loaded on
> mudballs
> startup the cmake-file class is not defined. Thus one needs to
> defclass
> cmake-file before the define-system.

Yes, this is currently the case. For all components we need to have
the class available, even if it is
just a stub.

> Is sysdef.cmake a good name? I will contact you for adding cl-smoke
> to the
> mudballs distribution when it's in a releasable state.

It sounds fine to me.

Cheers and thanks for the feedback,
Sean.

[Sean Ross]

On 18 Apr 2009, at 20:55, Paul Onions wrote:

On Apr 16, 7:48 pm, sross <ros...@gmail.com> wrote:

 Out of curiosity, with a show of hands, how many people are currently

using Mudballs on a daily basis?

I use Mudballs in a small way at work and on a personal project I'm
working on.

Glad to hear it.

I'm really only a dabbler in Lisp at the moment (time doesn't permit
much more than that), but the only reason I've found to step outside
the Mudballs environment is to install CLOCC, and that was just to get
access to the CLLIB/math routines (specifically an implementation of
the ERF function). It would be nice if I could get this in Mudballs,
but:
1. CLOCC has its own build system and porting this to Mudballs looks
decidedly non-trivial to me.

It does but we should be able to add CLLIB as  a system which mudballs can manage which should 

be a relatively trivial process. I'll add this to the list of systems to define.

I would be willing to port a simple (open) math library to Mudballs
for others to use, but I don't think I'm expert enough at Lisp yet to
attempt this with CLOCC.

I've just done a quick scan of the source and it would require some modification to work around CLOCC's 

approach of using require to load dependencies.

Cheers,

 Sean.

[Tobias Rautenkranz]

> > - There is no simple way to have mudballs search the current
> > directory first?
> > Solved this by loading the .mbd file with LOAD.
> > (see: http://tobias.rautenkranz.ch/lisp/cl-smoke/qt.tests/test.lisp)
> > I need this to run mb:test on darcs record.
>
> Let me give this some thought as I'm not quite sure what the correct
> approach is here.
> My understanding is that you want to load a system which, rather than
> being identified by
> version, is identified by location, is this correct?

Yes -- I want the system in the current directory to shadow the normal one.

I use this to run mb:test on the committed version; works like this:
1. I do a "darcs record" -> Darcs commits the changes.
2. Darcs creates a copy of the repository in a temporary directory.
3. Darcs runs the configured test command. I my case: "sh ./test.lisp"
4. When the command returns an error, the commit is reverted and otherwise the
commit is accepted.

[Sean Ross]

On 19 Apr 2009, at 19:10, Tobias Rautenkranz wrote:
> Yes -- I want the system in the current directory to shadow the
> normal one.
>
> I use this to run mb:test on the committed version; works like this:
> 1. I do a "darcs record" -> Darcs commits the changes.
> 2. Darcs creates a copy of the repository in a temporary directory.
> 3. Darcs runs the configured test command. I my case: "sh ./test.lisp"
> 4. When the command returns an error, the commit is reverted and
> otherwise the
> commit is accepted.
>

Indeed, I've raised a bug to address this (http://redmine.mudballs.com/issues/show/151
)

Thanks for all your input.

sean.