On Thursday, 12 October 2017 at 14:20:42 UTC, b4s1L3 wrote:
On Wednesday, 11 October 2017 at 13:38:10 UTC, b4s1L3 wrote:
On Wednesday, 11 October 2017 at 13:08:38 UTC, Petar Kirov [ZombineDev] wrote:
On Wednesday, 11 October 2017 at 10:20:01 UTC, b4s1L3 wrote:
On Monday, 9 October 2017 at 19:44:24 UTC, Martin Nowak wrote:
Glad to announce D v2.076.1.


This point release fixes a few issues over v2.076.1, see the changelog for more details.


- -Martin

I didn't test the beta and missed this regression discovered by a CRON at TravisCI:


Sound like we need more test coverage on the Project Tester (https://github.com/dlang/ci/blob/master/pipeline.groovy#L319).

The regression was discovered in iz, which was at least another time "victim" of a regression (almost day for day 1 year ago). https://github.com/BBasile/iz.

This time it is trivial to overcome. It happens in a contract so checking for a version number does the job and the release version is not affected at all.

I can still add dmd beta as D Compiler in the CI. I think that iz occasionally breaks because it covers many programming style, not because of its "edginess", which means that's its probably not worth adding it to the Project tester.

Also i'd like to say that the policy that is that regression fixes are commited on stable and that the fact that they only come to master in a "sync operation" is a problem.

In the travis yaml we have to test dmd, dmd beta (stable, not yet released) and finally dmd master (current working tree). The policy should be changed so that regression fixes are commited to both master and stable, allowing to decrease the CI complexity.

The problem is mainly that testing a project with dmd beta is pointless. It's only useful 1 or 2 weeks before a release, which happens , let's say, 4 times per years, leading to a waste of computing resources at the CI service.

With reg fixes put in master at the same time that in stable, testing 3 versions of the DMD compiler would not be necessary anymore, i think.

Reply via email to