[GRASS5] A comment and a question.

GRASS GRASS developer
1

Folks,

It took a few days, but Hal Mueller convinced me that it probably is a
good idea for a "production" installation of GRASS 5.0 for the installed
files to be owned by root. That wasn't necessarily true for the beta
and pre-release versions, and we're certainly not talking about the
GRASSS database. For a stable system the installation scripts should
probably have an option to make root the owner of the installed files,
as are (for instance) the program files in a Postgres installation.
Last week Hal provided the list with some specific suggestions about how
that might be done.

As to the question... I seem to recall that my first GRASS 5
installation included an s.surf.krige program, but it isn't there in
either of my current installations. Is my memory completely messed up,
or was it removed? I searched the archives and found a couple letters
that mentioned kriging, but no explanation for what happened to the old
program. What happened to the old one, and are there any plans to
replace it?

Roger Miller
Lee Wilson and Associates

2

On Wed, 24 Oct 2001 20:53:30 -0600, "Roger S. Miller" <rgrmill@rt66.com> wrote:

Folks,

It took a few days, but Hal Mueller convinced me that it probably is a
good idea for a "production" installation of GRASS 5.0 for the installed
files to be owned by root. That wasn't necessarily true for the beta
and pre-release versions, and we're certainly not talking about the
GRASSS database. For a stable system the installation scripts should
probably have an option to make root the owner of the installed files,
as are (for instance) the program files in a Postgres installation.
Last week Hal provided the list with some specific suggestions about how
that might be done.

If someone wants to look at using "install" rather than tar, it will
do the right thing (and probably be faster than tar anyway). The
"install" program can also strip symbols, which could be useful
since initial binaries could have debugging symbols, and "install"
would remove them. Then a recompile wouldn't be necessary to grab
a debuggable version of a module.

As to the question... I seem to recall that my first GRASS 5
installation included an s.surf.krige program, but it isn't there in
either of my current installations. Is my memory completely messed up,
or was it removed? I searched the archives and found a couple letters
that mentioned kriging, but no explanation for what happened to the old
program. What happened to the old one, and are there any plans to
replace it?

It was/is s.surf.krig. Bugs ... from old copy of grass CVS:

[RC]s.surf.krig
  Not working. Needs various bugfixes.
  Calls to G_distance() seem to suffer stack corruption. The values
  of parameters passed in and out of G_distance() are corrupted.
  (being worked on by David D Gray)

--
Eric G. Miller <egm2@jps.net>

3

On Wed, Oct 24, 2001 at 09:07:54PM -0700, Eric G. Miller wrote:

On Wed, 24 Oct 2001 20:53:30 -0600, "Roger S. Miller" <rgrmill@rt66.com> wrote:

>
> Folks,
>
> It took a few days, but Hal Mueller convinced me that it probably is a
> good idea for a "production" installation of GRASS 5.0 for the installed
> files to be owned by root. That wasn't necessarily true for the beta
> and pre-release versions, and we're certainly not talking about the
> GRASSS database. For a stable system the installation scripts should
> probably have an option to make root the owner of the installed files,
> as are (for instance) the program files in a Postgres installation.
> Last week Hal provided the list with some specific suggestions about how
> that might be done.

If someone wants to look at using "install" rather than tar, it will
do the right thing (and probably be faster than tar anyway). The
"install" program can also strip symbols, which could be useful
since initial binaries could have debugging symbols, and "install"
would remove them. Then a recompile wouldn't be necessary to grab
a debuggable version of a module.

Perhaps we should start to think about library re-organization now,
to address (for 5.1)
- cleanup of mixed libraries
- extract the IO routines to support libgrassio (Frank Warmerdam)
- support shared libraries
- add the new makefile system (package is in grass51/ in CVS)

and keeping 5.0 as is for this issue.

> As to the question... I seem to recall that my first GRASS 5
> installation included an s.surf.krige program, but it isn't there in
> either of my current installations. Is my memory completely messed up,
> or was it removed? I searched the archives and found a couple letters
> that mentioned kriging, but no explanation for what happened to the old
> program. What happened to the old one, and are there any plans to
> replace it?

It was/is s.surf.krig. Bugs ... from old copy of grass CVS:

[RC]s.surf.krig
  Not working. Needs various bugfixes.
  Calls to G_distance() seem to suffer stack corruption. The values
  of parameters passed in and out of G_distance() are corrupted.
  (being worked on by David D Gray)

Since David can't support GRASS these days due to other business,
the GSTAT or R are candidates for kriging. Both support GRASS
directly.

Best regards

Markus

4

Eric G. Miller wrote:

> It took a few days, but Hal Mueller convinced me that it probably is a
> good idea for a "production" installation of GRASS 5.0 for the installed
> files to be owned by root. That wasn't necessarily true for the beta
> and pre-release versions, and we're certainly not talking about the
> GRASSS database. For a stable system the installation scripts should
> probably have an option to make root the owner of the installed files,
> as are (for instance) the program files in a Postgres installation.
> Last week Hal provided the list with some specific suggestions about how
> that might be done.

If someone wants to look at using "install" rather than tar, it will
do the right thing (and probably be faster than tar anyway).

Unfortunately, switching to "install" would be an involved task, which
would best be done as part of the new build system (i.e. in 5.1).

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

5

At 4:18 PM +0100 10/25/01, Glynn Clements wrote:

Unfortunately, switching to "install" would be an involved task, which
would best be done as part of the new build system (i.e. in 5.1).

"tar xfo <yournamehere>" would do the job also.

Minimally intrusive change for 5.0 would be an option to "chown -R ugo <yournamehere>" the tree after the install is done. No mods to stuff that already works, easy to leave out if it breaks. This wouldn't even have to be automated--could be just a mention in the install directions that root doesn't own the installed files and if it matters to you here's the command to fix it.

Note that an active developer, or someone who updates his installation frequently, probably would _not_ want root to own the installed tree.

I'm head down on a deadline at the moment, won't be able to contribute/respond for a few days.

Hal

6

Hal Mueller wrote:

> Unfortunately, switching to "install" would be an involved task, which
> would best be done as part of the new build system (i.e. in 5.1).

"tar xfo <yournamehere>" would do the job also.

For which tar? For GNU tar, the "o" switch is described as:

  -o, --old-archive, --portability write a V7 format archive

Minimally intrusive change for 5.0 would be an option to "chown -R
ugo <yournamehere>" the tree after the install is done.

That seems plausible; it should work so long as the user running "make
install" is either root or whoever ran "make". But how to get the
username? "su" won't change $USER.

It would probably better to just attempt "chown -R root ...".

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