[GRASS5] Update: 5.0.3-cvs -> 5.3.0-cvs

Hi,

I have modified the VERSION file in CVS:

5.0.3-cvs -> 5.3.0-cvs

The current (and desired) situation is now:
We have

- 5.0.3 release_branch (RC3 at time, waiting for release)
- 5.3.0-cvs: new PROJ engine, modified NVIZ, G3D updated
              tons of fixes
- 5.7.0-cvs: former 5.1, new vector engine, DBMS support etc

The 5.7.0-cvs instructions should refer to 5.3.0, I tried to
update the relevant texts such as INSTALL etc.

Please make further updates if needed.

After getting 5.0.3 out of the door we should consider a
first 5.3.0 and also 5.7.0 release.

Markus

The current (and desired) situation is now:
We have

- 5.0.3 release_branch (RC3 at time, waiting for release)
- 5.3.0-cvs: new PROJ engine, modified NVIZ, G3D updated
              tons of fixes
- 5.7.0-cvs: former 5.1, new vector engine, DBMS support etc

The 5.7.0-cvs instructions should refer to 5.3.0, I tried to
update the relevant texts such as INSTALL etc.

Please make further updates if needed.

After getting 5.0.3 out of the door we should consider a
first 5.3.0 and also 5.7.0 release.

I apologize for bringing up more meaningless version number debates,
but why 5.3.0 and not 5.3.9? Why not 5.4.0pre1? 5.3.0 seems many
releases away from 5.4.0 (even though the distance is meaningless).

Is 5.4.0 really that far off? What's missing besides testing, testing,
testing? Is there a list of critical bugs that need to be fixed before
release? (e.g. NVIZ/Tk 8.4 on some systems; OSX & solaris build
problems; r.terraflow CVS update) Is there any major missing
code or subsystems?

By the same argument, what's missing before a 5.7.0-experimental
release and grass51 CVS moving to 5.7.1? (yes, code merge- wait just
until 5.4.0 is out & thus in buxfixe only mode?) Is it too much to
unwind the 'make mix' symlinks and make a source-only 5.7.0 release now?
(I guess that means no CVS checkout of 5.7.0 and not a very good idea)

If we do a merge early before, 5.4.0 is out, applying changes to both
trees is double the work, a pain, and bug prone. The sooner 5.4.0 is out
and all new changes go to the 5.7 branch the better, I think.

just some humble thoughts,

Hamish

On Sat, Sep 13, 2003 at 01:24:19AM +1200, Hamish wrote:

> The current (and desired) situation is now:
> We have
>
> - 5.0.3 release_branch (RC3 at time, waiting for release)
> - 5.3.0-cvs: new PROJ engine, modified NVIZ, G3D updated
> tons of fixes
> - 5.7.0-cvs: former 5.1, new vector engine, DBMS support etc
>
> The 5.7.0-cvs instructions should refer to 5.3.0, I tried to
> update the relevant texts such as INSTALL etc.
>
> Please make further updates if needed.
>
> After getting 5.0.3 out of the door we should consider a
> first 5.3.0 and also 5.7.0 release.

I apologize for bringing up more meaningless version number debates,
but why 5.3.0 and not 5.3.9? Why not 5.4.0pre1? 5.3.0 seems many
releases away from 5.4.0 (even though the distance is meaningless).

I don't think that it is releases away.

Is 5.4.0 really that far off? What's missing besides testing, testing,
testing? Is there a list of critical bugs that need to be fixed before
release? (e.g. NVIZ/Tk 8.4 on some systems; OSX & solaris build
problems; r.terraflow CVS update) Is there any major missing
code or subsystems?

The latest r.terraflow code I received yesterday via Helena.
However, I hesitate to just copy over in CVS. Is there anyone
interested to try it first and then upload into CVS?
For the moment, the new src code is here:

http://mpa.itc.it/markus/tmp/r.terraflow.1.5.tar.gz

By the same argument, what's missing before a 5.7.0-experimental
release and grass51 CVS moving to 5.7.1? (yes, code merge- wait just
until 5.4.0 is out & thus in buxfixe only mode?) Is it too much to
unwind the 'make mix' symlinks and make a source-only 5.7.0 release now?
(I guess that means no CVS checkout of 5.7.0 and not a very good idea)

Right now I have added a 'srcdist' target which copies the linked 5.3.x files
into a 5.7.0-cvs src tarball:

make srcdist

and you are there.

If we do a merge early before, 5.4.0 is out, applying changes to both
trees is double the work, a pain, and bug prone. The sooner 5.4.0 is out
and all new changes go to the 5.7 branch the better, I think.

I agree. And mainly motivated by the hope, that more developers switch
over to the 5.7 branch.

Best

Markus

On Sat, Sep 13, 2003 at 01:24:19AM +1200, Hamish wrote:

> The current (and desired) situation is now:
> We have
>
> - 5.0.3 release_branch (RC3 at time, waiting for release)
> - 5.3.0-cvs: new PROJ engine, modified NVIZ, G3D updated
> tons of fixes
> - 5.7.0-cvs: former 5.1, new vector engine, DBMS support etc
>
> The 5.7.0-cvs instructions should refer to 5.3.0, I tried to
> update the relevant texts such as INSTALL etc.
>
> Please make further updates if needed.
>
> After getting 5.0.3 out of the door we should consider a
> first 5.3.0 and also 5.7.0 release.

I apologize for bringing up more meaningless version number debates,
but why 5.3.0 and not 5.3.9? Why not 5.4.0pre1? 5.3.0 seems many
releases away from 5.4.0 (even though the distance is meaningless).

After 5.3.9 comes 5.3.10, so that would not help much ... :slight_smile:
We actually want to have more experimental releases more directly
from the CVS tree. So it is fine to assume we might have 5.0.4.
5.3.1 5.3.2 and a few more.

Is 5.4.0 really that far off? What's missing besides testing, testing,
testing?

We don't know, only testing can tell,
but according to software engineering that takes up more then 50%
of the actually software construct efforts, so yes I believe 5.4.0
is a bit away.

Is there a list of critical bugs that need to be fixed before
release? (e.g. NVIZ/Tk 8.4 on some systems; OSX & solaris build
problems; r.terraflow CVS update) Is there any major missing
code or subsystems?

We should put bugs in the bugtracker
and quite some of them are in there.
So they probably need categorisation.

By the same argument, what's missing before a 5.7.0-experimental
release and grass51 CVS moving to 5.7.1? (yes, code merge- wait just
until 5.4.0 is out & thus in buxfixe only mode?) Is it too much to
unwind the 'make mix' symlinks and make a source-only 5.7.0 release now?
(I guess that means no CVS checkout of 5.7.0 and not a very good idea)

There was a plan once, before the new roadmap came out.
Dig in the mailinglist archives.

If we do a merge early before, 5.4.0 is out, applying changes to both
trees is double the work, a pain, and bug prone. The sooner 5.4.0 is out
and all new changes go to the 5.7 branch the better, I think.

Yes, but there never was enough help to actually control other
developers to not add non-critical bug fixes to the other branches.

The latest r.terraflow code I received yesterday via Helena.
However, I hesitate to just copy over in CVS. Is there anyone
interested to try it first and then upload into CVS?
For the moment, the new src code is here:

http://mpa.itc.it/markus/tmp/r.terraflow.1.5.tar.gz

I don't know enough about r.terraflow, C++, or the build system to be
useful here, but here is the diff for those who do.

(attachments)

r.terraflow-jan_apr.diff (16.1 KB)

Hamish wrote:

> The latest r.terraflow code I received yesterday via Helena.
> However, I hesitate to just copy over in CVS. Is there anyone
> interested to try it first and then upload into CVS?
> For the moment, the new src code is here:
>
> http://mpa.itc.it/markus/tmp/r.terraflow.1.5.tar.gz

I don't know enough about r.terraflow, C++, or the build system to be
useful here, but here is the diff for those who do.

Don't update the Gmakefile; the one in CVS is correct (well, it isn't
correct, but it's an improvement over the updated version).

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