[GRASS5] GRASS 5.0.0stable to be released soonish

[Glynn Clements]

libgis/Init.sh I18N:

I suggest no for final, but maybe if we decide to do a pre5. The problem
is that, however trivial the changes might appear, if something does go
wrong, the whole of GRASS may be completely unusable.

> > src/libes/proj/PJ_krovak.c
> > src/libes/proj/PJ_krovakgis.c
> > 	krovak[gis] addition
> 
> This is pretty safe. It should appear in PROJ4 soon (hi Frank).
> Of course it can also be skipped.

I can't see any way in which this can break anything.

> > html/Gmakefile
> > 	[ -> test
> > 	mkdir -> mkdir -p
> 
> yes, needed

The autoconf docs specifically warn against using "mkdir -p", as it
isn't supported on all platforms. However, GRASS already relies upon
it in a number of places.

> > 	fix postgresql.org URL
> 
> depends

On what? The actual change is:

-This program requires the <a href="www.postgresql.org">PostgreSQL</a> database
+This program requires the <a href="http://www.postgresql.org">PostgreSQL</a> database

Although, it should really be:

	http://www.postgresql.org/

URLs which refer to directories (i.e. an implicit "index.html") should
always have a trailing slash. Otherwise, the HTTP server has to reply
with a redirect (301/302) response rather than the actual file. This
is so that relative URLs work correctly.

However, I suspect that the browser will fix this automatically in the
case of a "top level" URL such as the above.

> > html/html/d.text.freetype.html
> > 	-rpsc flags
> > 	align=, rotation=
> > 	Command mode
> 
> no idea

> > src/display/d.text.freetype/Gmakefile
> > src/display/d.text.freetype/freetypecap
> > src/display/d.text.freetype/main.c
> > 	add align=, rotation= options
> > 	command mode
> 
> no idea: new feature

These are enhancements rather than bug fixes. However, d.text.freetype
is relatively new, so backward compatibility isn't really an issue yet.

> > mk/mid.mk
> > mk/mid.mk.shlib
> > 	add LOCALE definition
> yes/depends ?

> > src/CMD/generic/make.mid
> > 	define LOCALE (currently unused)
> 

These aren't used by anything presently, so it doesn't really matter
either way.

> > * src/CMD/generic/GISGEN.sh
> > 	"locale" subdirectory
> 
> depends

This is a definite yes. This was the primary cause of the "make install"
problem in pre5.

[long list of files snipped]
> > 	various tidying:
> > 		include necessary headers
> > 		remove unused variables
> > 		error handling changes
> > 		use $GRASS_PAGER instead of $PAGER
> > 		remove $Log$
> 
> yes

Then the next version should be labelled "pre5". 

> > src/display/devices/PNGdriver/Graph_Set.c
> > 	Fix GRASS_BACKGROUNDCOLOR processing
> 
> seems to be needed?

Yes.

> > src/display/devices/PNGdriver/README
> > 	default background is black
> yes if above

Yes regardless; the default background *is* black, but the documentation
said it was white. This just fixes the documentation to match the actual
behaviour.

> > src/display/devices/XDRIVER/XDRIVER24/Graph_Set.c
> > 	add "GRASS" to monitor title (tcltkgrass bug workaround)
> 
> seems to be needed?

The tcltkgrass "stop all X monitors" option relies upon the XDRIVER
window title containing the string "GRASS". The tcltkgrass code is an
ugly hack, which should ultimately be fixed; but this should suffice for
now.

> > src/general/init/Gmakefile
> > 	generate $GISBASE/etc/license
> > 	link $GISBASE/etc/lock against libgis
> 
> yes

The first of these is part of the Init.sh I18N.

The second of these relates to the error handling changes. etc/lock now
uses G_fatal_error(), so it has to be linked against libgis.

However, I have no idea how this interacts with the fact that etc/lock
is run during GRASS initialisation. etc/lock doesn't call G_gisinit()
(which may not even work at that point); I don't know whether that
presents a problem for the use of G_fatal_error().

> > src/general/init/init.sh
> > *	add "savehist" setting to .cshrc
> yes (trivial fix)
> 
> > 	i18n
> no idea

> > 	licence notice moved to separate file
> 
> mhh, if it helps

Part of Init.sh I18N. The idea is that the licence and introductory
message are read from separate (localised) files.

> > src/mapdev/v.proj/Gmakefile
> > * src/mapdev/v.proj/main.c
> > * src/sites/s.proj/main.c
> > 	datum/ellipse fixes
> 
> I suggest that then also m.proj[2] must be updated to
> avoid inconsistencies

OK; I'll look into it.

> > src/raster/r.proj/cmd/bordwalk.c
> > 	ignore invalid points
> 
> no idea (Morten?)

This part is definitely recommended.

> > src/raster/r.proj/cmd/main.c
> > src/raster/r.proj/cmd/r.proj.h
> > *	datum/ellipse fixes
> I suggest that then also m.proj[2] must be updated to
> avoid inconsistencies
> 
> > 	tidy
> > 	add -n (nocrop) flag
> > 	move "menu" array from r.proj.h to main.c
> 
> maybe yes?

I'm fairly sure about the "-n" switch, but we should probably re-enable
the other call to bordwalk(), also on that switch. But:

1. This needs to be tested, preferably with a wide variety of
projections

2. We need to consider whether cropping should be on or off by default;
enabling cropping saves memory, but could result in the output being
incomplete.

3. At some point, we should consider how to handle latitudinal wrapping.

> > src/tcltkgrass/main/gis_set.tcl
> > src/tcltkgrass/tcltkgrass
> > 	I18N
> 
> no idea

This could still do with some more work (e.g. checking LC_MESSAGES,
LC_ALL). Not as critical as libgis or Init.sh, but still fairly
important to get right.

> > Bogus Diffs
> > ===========
> > 
> > src/libes/gis/sites.c
> > src.contrib/GMSL/NVIZ2.2/src/togl.c
> 
> What means Bogus here?

It means that these files only differ due to the comments inserted by
$Log$, so they aren't really different.

Summary: even if we omit any changes which are enhancements rather than
bug fixes, there are still enough changes that we should seriously
consider releasing a pre5 version for testing.

-- 
Glynn Clements <glynn.clements at virgin.net>