Am Donnerstag, dem 03.12.2020 um 12:33 +0000 schrieb James: > On 03/12/2020 11:46, Michael Käppler wrote: > > What do you mean with "full reg test suite"? Additional examples that > > are not in the normal reg test suite? > > Well in the 'olden days' we'd have a set of tests on the website - here > is an old URL that no longer works > (http://lilypond.org/test/v2.13.51-1/compare-v2.13.50-1/index.html > <http://lilypond.org/test/v2.13.51-1/compare-v2.13.50-1/index.html>) - > and as this was autogenerated as part of the website build, it was quick > and easy to see if there were any changes that had altered between, say, > last unstable or between current unstable and previous stable.
This continues to be uploaded by GUB to http://lilypond.org/test/, e.g. http://lilypond.org/test/v2.21.81-1/compare-v2.21.80-1/index.html > > > I think comparisons between several versions are of limited usefulness, > > because it can be difficult to see if > > changes are intended, not intended, or even occuring at random. > > Quite, but if you do this 'check' as part of the workflow for the next > official release (Phil may already, I don't know and I don't want to > necessarily make more work for anyone) then at least we aren't in for > any shocks and also, anything that is intended should be in the > changes.tely file, so this also helps us keep that up to date. But I am > not sure how complex this is to set up again and if it is worth the work > or not. ... and as it already exists, I am not sure how many looks it will get (I look at the results a semi-regularly, but as Michael noted something like changing the default font or tweaking horizontal distancing makes the output pretty useless). Jonas > > > In my opinion the current test suite has conceptual problems that go > > beyond that discussion. > > E.g. that we don't compare actual output, but grob positions and > > extents, AFAIK. > > That lead to problems in Harm's MR: > > https://gitlab.com/lilypond/lilypond/-/merge_requests/497 > > Also issues like https://gitlab.com/lilypond/lilypond/-/issues/720 that > > affect specific backends cannot be > > tested (or it would require hacks) > > Yes that is beyond the scope of what I am talking about, but > fundamentally it might be useful to someone ('perfect enemy of good' and > all that). > > > James >
signature.asc
Description: This is a digitally signed message part
