Re: New gnucap development snapshot

On Sunday 09 July 2006 17:47, Dan McMahill wrote:
> al davis wrote:
> > There is a new development snapshot of gnucap available.
> >
> >
> >
> > You can get it at:
> > Official site:
> > http://www.gnucap.org/devel/gnucap-2006-07-08.tar.gz
> > Mirrors:
> > http://www.geda.seul.org/dist/gnucap-2006-07-08.tar.gz
>
> cool.  I was updating the NetBSD package and ran into 3 minor
> things that I'm somewhat puzzled by.
>
> Did you do anything to the tar file after running 'make
> dist'? 

This is a transitional release.  The tar file was actually made 
by the old system.  The new system didn't work correctly, and 
rather than taking the time to troubleshoot, I fell back to the 
old.

Actually, I spent quite a bit of time troubleshooting in 
general.  I don't yet fully understand what the issues are.

> I'm asking because the man/gnucap-man.dvi file 
> doesn't seem to be in the tar file but as near as I can tell
> from man/Makefile.am, it should be.  

It wasn't in the old dist file.  I consider it to 
be "non-source" and those who don't have TeX installed can't 
use it anyway.  It is an intermediate file.

> The result is 'make 
> install' fails if I don't have latex installed (the build
> system is set up to use latex if found but otherwise issue
> warnings and use the .dvi file which ships in the .tar.gz).

Maybe that explains some of the behavior I was seeing.  The 
system I use mostly las latex installed.

> The second question is if your intention is to not have the
> html manual installed with 'make install'.  By commenting out
> the SUBDIRS= html line in man/Makefile.am, the html manual
> doesn't get installed.  Whats wierd is by commenting that
> out, the html manual shouldn't have even ended up in the
> .tar.gz file yet I see it there.

As I said, the tar.gz file was made the old way, which does 
include the html manual.  Even though it is non-source, for 
many users, it is a preferred way to view the docs.  Both html 
and pdf are provided, so you can read the manual right away 
without building anything.  Should this be changed?  I wonder, 
because both of these are non-source.

> My last question is really an observation that
> test/==/Makefile.in seems to have been generated by a really
> old automake (1.4) while all the others are from a modern
> (1.9.6) automake.  I'm not sure how this one got singled out.

So that explains it ......

That whole directory is generated.  Same goes for man/html.  So, 
when tinkering, I might remove it then rebuild.  The rebuild 
didn't put in Makefile.am, so Makefile.in wasn't there either, 
so I put it in manually.

> I can try to investigate all of these but don't want to spend
> time on it if the .tar.gz file was modified after creation
> with 'make dist' or (better yet) 'make distcheck'.

As I said, I used the old script.

When I make a major change, I try to make sure the new way is 
fully functional before removing the old one.  This means that 
there is usually a time when both systems are there, and both 
being maintained.

Sometimes, I can make the transition in one snapshot.  
Sometimes, it may even span major releases.  Look 
at "named-nodes" as an example of one that took much longer 
than it should have.  There was at least one snapshot where I 
didn't even tell anyone it was there, until somebody tried it.  
For a while, both were there, and it defaulted to numbers.  For 
a while, both were there, and it defaulted to names.  Then the 
numbers went out and they are always named.  For the 
transitional versions, it wasn't fully working.

I have a lot to say about this snapshot.  Mostly it is about the 
new time step control.  Prepare for some reading!