Ahmon Dancy | 4 Aug 2004 01:08
Picon
Favicon

Re: Patch for Allegro CL module

I committed some additional changes today to help out w/ case mode
issues.

William S Fulton | 4 Aug 2004 23:59
Picon
Favicon
Gravatar

SWIG pdf and single page HTML documentation

I've spent some time polishing up the SWIG documentation. My goal was to 
generate the documentation into pdf format. To achieve this I've used a program 
called htmldoc which generates a pdf document from html files. This program can 
also generate generate a single html page from many html pages. This is great as 
many web sites offer their html documentation in smaller bite sized pieces as 
well as one huge file and so can we now.

I've had to fix many hundreds of html errors in order for all this work. The 
html validates now using the html tidy (sometimes just called tidy) program from 
http:://www.w3.org. It is a commonly available package on Linux. The Doc/Manual 
directory now has a makefile which can be used by SWIG developers to generate 
the pdf and single page html documentation. It requires tidy and and htmldoc to 
be installed. I've also fixed the makechap.py script to generate valid html too. 
It is still used to generate the table of contents for the 'multiple page' html 
documentation. htmldoc produces a really neat TOC for the pdf format too.

I've had to work around a couple of htmldoc limitations. The first is the 
anchors (<a name="somename">) in each file are generated into a global 
namespace. So all anchors must be uniquely named across all the .html files. 
This means that the n1, n2, n3 anchor names that makechap.py was replicating in 
each html file were no good. Now it genrates a name based on the html file name, 
eg all anchors in Python.html are called Python_nn1, Python_nn2 etc. Any of 
these autogenerated anchor names can be replaced by another name. However if 
someone changes the anchor name to something else then that name will be kept 
and not overidden by makechap.py. This makes it possible to use references to 
headings which will not be altered by future invocations of makechap.py. One 
other limitation of htmldoc is making links to just a filename, say
<a href="SWIG.html">. For the single page html doc generation, links must refer 
to an anchor name, not just the filename, eg use <a href="SWIG.html#SWIG"> instead.

(Continue reading)

William S Fulton | 11 Aug 2004 23:29
Picon
Favicon
Gravatar

SWIG-1.3.22 release

A SWIG release is long overdue. Any objections if I make a release of the CVS 
version? I can do the whole release and have the week beginning 23 Aug in mind.

William

William A. Hoffman | 12 Aug 2004 04:41
Picon

Re: SWIG-1.3.22 release

Can you make sure that my patches in the bug tracker are applied or at least evaluated?
There are some serious problems with multiple module tcl wrappings where import is used.

-Bill

At 05:29 PM 8/11/2004, William S Fulton wrote:
>A SWIG release is long overdue. Any objections if I make a release of the CVS version? I can do the whole
release and have the week beginning 23 Aug in mind.
>
>William
>_______________________________________________
>Swig-dev mailing list  -  Swig-dev <at> cs.uchicago.edu
>http://mailman.cs.uchicago.edu/mailman/listinfo/swig-dev

William A. Hoffman | 12 Aug 2004 16:07
Picon

Re: SWIG-1.3.22 release

Here is the bug id:

981362 

Looks like I did not submit the patch to the bug tracker,
but I did post it to the list:
http://mailman.cs.uchicago.edu/pipermail/swig-dev/2004-June/014000.html

The version of swig that I made the changes to is a bit old.
However, the basic idea is sound.   

Change the swig_class struct for tcl to use
a const char**bases  instead of pointing to swig_class*.
Then change all the code that uses the bases pointer 
to get the class this way:

(swig_class *) SWIG_TypeQuery(base_name)->clientdata;

That way the SWIG_TypeQuery is not called until the interpreter is
running code. The way it is now SWIG_TypeQuery is called during initialization
and it may or may not find imported classes depending on the order that
they are loaded into tcl.  A class may even end up with itself as a base
class....

diff -x CVS -r Source/Swig/typesys.c /cygdrive/c/hoffman/My Builds/Insight/Utilities/CableSwig/SWIG/Source/Swig/typesys.c

1448,1449c1441,1446
<       while (rk.key) {
<         clientdata = SwigType_clientdata_collect(rk.key,ch);
---
(Continue reading)

Salvador "Fandiño" | 12 Aug 2004 16:15
Picon
Favicon
Gravatar

Re: SWIG-1.3.22 release

I (sfandino on SF) have also submited a couple of patches for some
important bugs on the Perl module...

- Salvador

--- "William A. Hoffman" <billlist <at> nycap.rr.com> wrote:

> Here is the bug id:
> 
> 981362 
> 
> 
> Looks like I did not submit the patch to the bug tracker,
> but I did post it to the list:
> 
> ...

		
__________________________________
Do you Yahoo!?
New and Improved Yahoo! Mail - Send 10MB messages!
http://promotions.yahoo.com/new_mail 

David Beazley | 12 Aug 2004 21:27
Favicon

SWIG-1.3.22 release

William S Fulton writes:
 > A SWIG release is long overdue. Any objections if I make a release of the CVS 
 > version? I can do the whole release and have the week beginning 23 Aug in mind.
 > 

Yep.  It's been awhile.  How stable is CVS?   Is the documentation up
to date?   Other than that, I'm ok with a release.

-- Dave

William S Fulton | 12 Aug 2004 22:19
Picon
Favicon
Gravatar

Re: SWIG-1.3.22 release

I'm going through some of the patches and am committing those that I feel 
comfortable with. The runtime system is not one of those areas, so hopefully 
Dave or Matthias Koeppe or John Lenz could take a look and commit.

William

William A. Hoffman wrote:
> Here is the bug id:
> 
> 981362 
> 
> 
> Looks like I did not submit the patch to the bug tracker,
> but I did post it to the list:
> http://mailman.cs.uchicago.edu/pipermail/swig-dev/2004-June/014000.html
> 
> 
> The version of swig that I made the changes to is a bit old.
> However, the basic idea is sound.   
> 
> Change the swig_class struct for tcl to use
> a const char**bases  instead of pointing to swig_class*.
> Then change all the code that uses the bases pointer 
> to get the class this way:
> 
> (swig_class *) SWIG_TypeQuery(base_name)->clientdata;
> 
> 
> That way the SWIG_TypeQuery is not called until the interpreter is
> running code. The way it is now SWIG_TypeQuery is called during initialization
(Continue reading)

William S Fulton | 12 Aug 2004 22:29
Picon
Favicon
Gravatar

Re: SWIG-1.3.22 release

Can some Perl experts look at this? I see you have also a patch for adding 
director support for Perl, however it is in a new Perl module. It really needs 
adding to the current Perl module in the same manner as the other modules that 
support directors.

William

Salvador Fandiño wrote:

> I (sfandino on SF) have also submited a couple of patches for some
> important bugs on the Perl module...
> 
> - Salvador
> 
> 
> --- "William A. Hoffman" <billlist <at> nycap.rr.com> wrote:
> 
> 
>>Here is the bug id:
>>
>>981362 
>>
>>
>>Looks like I did not submit the patch to the bug tracker,
>>but I did post it to the list:
>>
>>...
> 
> 
> 
(Continue reading)

William S Fulton | 12 Aug 2004 22:35
Picon
Favicon
Gravatar

Re: SWIG-1.3.22 release

David Beazley wrote:

> William S Fulton writes:
>  > A SWIG release is long overdue. Any objections if I make a release of the CVS 
>  > version? I can do the whole release and have the week beginning 23 Aug in mind.
>  > 
> 
> Yep.  It's been awhile.  How stable is CVS?   Is the documentation up
> to date?   Other than that, I'm ok with a release.
> 
CVS is quite stable. I'll move any broken testcase that I can't fix to the set 
of broken testcases, although I don't see many broken testcases.

I've been adding documentation to stuff I've been doing and have done the pdf 
docs now. Marcelo is the other person who has done quite a bit since the last 
release, but am not sure whether any of that needs documenting.

William


Gmane