I also don’t know. It is an addition or feature, so it should not (doesn’t have to) go to release. It is nor really mature either, although it apparently runs and works. The distribution is still not solved. Will the tests be included in installation? What about compiled C test programs (now compiled and added to standard modules)? So far, it has been tested just on Ubuntu as far as I know. (More discussion on that should be on GSoC project page)
From the point of view of testing, it would be good to have it in release branch as soon as possible, so I don’t know.
Maybe it can go to GRASS GIS 7.0.1 but not to 7.0.0.
On 11 November 2014 23:32, Vaclav Petras <wenzeslaus@gmail.com> wrote:
On Tue, Nov 11, 2014 at 5:05 PM, Luca Delucchi <lucadeluge@gmail.com> wrote:
I don't understand if gunittest will be backported before 7.0 release,
if we backport we can leave that line, otherwise it should be removed
I also don't know. It is an addition or feature, so it should not (doesn't
have to) go to release. It is nor really mature either, although it
apparently runs and works. The distribution is still not solved. Will the
tests be included in installation? What about compiled C test programs (now
compiled and added to standard modules)? So far, it has been tested just on
Ubuntu as far as I know. (More discussion on that should be on GSoC project
page)
I think we should have a "make test" to run the tests.
I tested it in Debian too
From the point of view of testing, it would be good to have it in release
branch as soon as possible, so I don't know.
Maybe it can go to GRASS GIS 7.0.1 but not to 7.0.0.
On Tue, Nov 11, 2014 at 11:32 PM, Vaclav Petras <wenzeslaus@gmail.com> wrote:
Maybe it can go to GRASS GIS 7.0.1 but not to 7.0.0.
Personally I think we should include the gunittest to grass 7.0.0.
It is just an extra functionality that does not affect GRASS
modules/GUI/libraries in any way.
So could be that the gunittest is not working... but most (nearly) all
the users won't notice.
Instead could be good to have the gunittest on grass70 to test in
other/more platforms (OS and distributions),
could be good to have a way from the GUI to launch the test and see the report.
Moreover we are a small team of developers, we can not do extensive
tests for nearly 3*10^6 lines of code [0], so I personally believe
that we should just stop to worry too much about stability and just
release every two/four months (using some dedicated branch with
limited scope, such as support for python3, review the raster API,
etc).
I'm using with my colleagues grass71 for the daily work and seems to
work without problems.
Hi Vaclav,
i think the testing framework should go into 7.0.0 directly. Since, as
Pietro mentioned, it will not harm GRASS users. IMHO the testing
framework is a key feature, very valuable, it works well and is
utterly useful to detect backport issues. Hence, using the testing
framework a new release policy may be to reduce test failures to zero
on Linux platforms.
Best regards
Soeren
2014-11-11 23:32 GMT+01:00 Vaclav Petras <wenzeslaus@gmail.com>:
On Tue, Nov 11, 2014 at 5:05 PM, Luca Delucchi <lucadeluge@gmail.com> wrote:
I don't understand if gunittest will be backported before 7.0 release,
if we backport we can leave that line, otherwise it should be removed
I also don't know. It is an addition or feature, so it should not (doesn't
have to) go to release. It is nor really mature either, although it
apparently runs and works. The distribution is still not solved. Will the
tests be included in installation? What about compiled C test programs (now
compiled and added to standard modules)? So far, it has been tested just on
Ubuntu as far as I know. (More discussion on that should be on GSoC project
page)
From the point of view of testing, it would be good to have it in release
branch as soon as possible, so I don't know.
Maybe it can go to GRASS GIS 7.0.1 but not to 7.0.0.
I forgot to mention this but incompatibility with 2.6 is another argument against backport. See the Tests down to 65% (from 86%) thread for further details.
However, nothing should stop you from adding gunittest to PYTHON_HOME and running trunk tests with release branch GRASS.
Hi Vaclav,
i think the testing framework should go into 7.0.0 directly. Since, as
Pietro mentioned, it will not harm GRASS users. IMHO the testing
framework is a key feature, very valuable, it works well and is
utterly useful to detect backport issues. Hence, using the testing
framework a new release policy may be to reduce test failures to zero
on Linux platforms.
I don’t understand if gunittest will be backported before 7.0 release,
if we backport we can leave that line, otherwise it should be removed
I also don’t know. It is an addition or feature, so it should not (doesn’t
have to) go to release. It is nor really mature either, although it
apparently runs and works. The distribution is still not solved. Will the
tests be included in installation? What about compiled C test programs (now
compiled and added to standard modules)? So far, it has been tested just on
Ubuntu as far as I know. (More discussion on that should be on GSoC project
page)
From the point of view of testing, it would be good to have it in release
branch as soon as possible, so I don’t know.
Maybe it can go to GRASS GIS 7.0.1 but not to 7.0.0.
On Fri, Jun 5, 2015 at 2:04 AM, Pietro <peter.zamb@gmail.com> wrote:
Hi Vaclav,
What do you think, is it time to move the gunittest suite from trunk to 70?
Hi Pietro,
I'm not sure. The API was not tested much especially the one for running
tests. Not that there would be some (serious) bugs. but I might
significantly improve the API/CLI usability after some feedback from other
people using it. The execution interface or its implementation might be
also changed with addition of `--exec` flag (#2579).
Perhaps the potential changes could be simply backported when testing
framework is in 7.0 branch, but still it is some additional work and I'm
not sure if it is worth since testing framework is still not used much (see
e.g. #2687 but also older large test failures or number of tested modules
-- < 7 %).