After another happy morning's fiddling with "make test" in the Stylesheets pureodd branch, I can confirm that: a) something is wrong in the way pure odd declarations are being generated for the tests which process ODD files b) there are a few changes in presence of linebreaks or whitespace, but they all seem to be benign c) otherwise all the tests work fine More precisely: a) I tweaked the Test/Makefile to suppress the diff commands (if you want to try this, update your version, and set REGENERATE to 1 before running "make") b) I then ran "make regenerate" to make a new directory containing "new-expected-results" and compared that to the existing "expected-results" c) Specifically: test15.odd.html and test34.odd. html look ok except that they have incorrect pure odd declarations (it's just empty). Curiously, in test-pure2.odd.html the pure odd decl is correct, but not nicely formatted test20.tex : both old and new versions generate apparently identical pdf mdtest1.tei includes some text not present in the existing expected-results; I suspect this is a genuine update d) There are several files in expected-results which I don't seem to have in my output from make test; I havent worked out why yet Hugh, can you spare a moment to look at this? I'm going to have LUNCH L 14/12/15 22:51, Martin Holmes wrote:
On 15-12-14 11:20 AM, Lou Burnard wrote:
I had a quick look at this, hoping that just adding -B to the diff commands (which ignores blank lines) would do the trick, but sadly no. Whatever has changed has the effect of introducing newlines at different places from those in expected-results, see e.g. test20.
I suppose one (sneaky) solution might be to modify the Makefile so as not to do any diffs, and then copy the results all to expected-results. This may sound absurd, but I am wondering how expected-results ever got created if not by some such trick.
Each test was originally created that way, I think, and they're usually modified that way when they change. If a diff -B of all the test results against expected-results gives no differences, I think you would be fine to copy them all to expected-results.
Cheers, Martin
On 14/12/15 18:39, Martin Holmes wrote:
There's a procession of tests which are failing because an edit on the pureodd branch introduced a blank line in the output which is not reflected in the expected-results. I've fixed a few of them, in the hope that this was an isolated case, but I'm now thinking that this is a general problem that should be fixed in the place that introduced the difference. I don't have time to track that down right now, but I think Hugh made those changes; Hugh, if you can figure out what's adding that blank line, you could revert the tree to b7ee9fa3d851dbd2ce3d560a1718b4aec40588f6, prior to my changes this morning, and fix the source rather than the symptoms.
Cheers, Martin
On 15-12-14 09:58 AM, Martin Holmes wrote:
The Stylesheets pureodd branch failed to build on Jenkins, which is why the P5 branch can't build. The Stylesheets build seems to be failing on:
diff test.html expected-results/test.html
The difference between them appears to be a single extra blank line at line 797 in test.html. I shall endeavour to fix that, and we'll see how we get on.
Cheers, Martin
On 15-12-14 09:22 AM, Lou Burnard wrote:
If we're reaching consensus on the meta-question of how issues should be labelled, could we now collectively consider at least some of the urgent outstanding issues themselves? In particular, I am still waiting for an explicit green light from at least one other council member who has successfully built using the lb42-pure-odd-2 branch? Martin set up a Jenkins job to do this, but it seems to have fallen over at the penultimate hurdle, and I don;'t know enough about Jenkins to see why. If we get that, then I could maybe issue a pull request (whatever that is) and get this stuff merged in...
If we're looking for jobs for new council members to cut their teeth on, I'd be delighted to propose trying to build this branch! Much more fun than fixing a tedious bit of prose.
I'd also appreciate some feedback on my ticket proposing a new anyXML element...
L