I don't get this. I just tried a local `make test` in Stylesheets (setting the DEFAULTSOURCE to my local 'dev' copy of p5subst), and it failed because of an expected results error. But I personally copied over the expected results file in question yesterday or the day before. The diff failure is: result: ! defined in <ref target="http://www.tei-c.org/release/doc/tei-p5-doc/en/html/BIB.html#XSLT2">6.2. </ref>.</p> exptected result: ! defined in <ptr target="#XSLT2"/>.</p> and 1/2 dozen others exactly like that. I.e., w/ full URI instead of just bare name identifier. What gives? Has anyone changed anything that should have anything to do with this? Is it something on my local system? (That's a rhetorical question.) Could one of the CI maintainers kick off the Stylesheets-dev-against-TEIP5-dev job?
I think it is something local, since at first blush it seems to work in the docker container. But if someone could kick off Stylesheets-dev-against-TEIP5-dev anyway, that would be good. (Or will it kick off automatically when TEIP5-dev is done?)
I don't get this. I just tried a local `make test` in Stylesheets (setting the DEFAULTSOURCE to my local 'dev' copy of p5subst), and it failed because of an expected results error. But I personally copied over the expected results file in question yesterday or the day before.
The diff failure is:
result: ! defined in <ref target="http://www.tei-c.org/release/doc/tei-p5-doc/en/html/BIB.html#XSLT2">6.2. </ref>.</p>
exptected result: ! defined in <ptr target="#XSLT2"/>.</p>
and 1/2 dozen others exactly like that. I.e., w/ full URI instead of just bare name identifier. What gives? Has anyone changed anything that should have anything to do with this?
Is it something on my local system? (That's a rhetorical question.) Could one of the CI maintainers kick off the Stylesheets-dev-against-TEIP5-dev job?
I'll kick mine off when the current P5-dev build is complete. Cheers, Martin On 2020-02-12 8:24 a.m., Syd Bauman wrote:
I think it is something local, since at first blush it seems to work in the docker container. But if someone could kick off Stylesheets-dev-against-TEIP5-dev anyway, that would be good. (Or will it kick off automatically when TEIP5-dev is done?)
I don't get this. I just tried a local `make test` in Stylesheets (setting the DEFAULTSOURCE to my local 'dev' copy of p5subst), and it failed because of an expected results error. But I personally copied over the expected results file in question yesterday or the day before.
The diff failure is:
result: ! defined in <ref target="http://www.tei-c.org/release/doc/tei-p5-doc/en/html/BIB.html#XSLT2">6.2. </ref>.</p>
exptected result: ! defined in <ptr target="#XSLT2"/>.</p>
and 1/2 dozen others exactly like that. I.e., w/ full URI instead of just bare name identifier. What gives? Has anyone changed anything that should have anything to do with this?
Is it something on my local system? (That's a rhetorical question.) Could one of the CI maintainers kick off the Stylesheets-dev-against-TEIP5-dev job?
Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
-- ------------------------------------------ Martin Holmes UVic Humanities Computing and Media Centre
My Jenkins fails to build both the dev branch and the new release branch: <https://teijenkins.hcmc.uvic.ca/job/Stylesheets-dev/42/parsed_console/> <https://teijenkins.hcmc.uvic.ca/job/Stylesheets-release-7.49.0/10/parsed_console/> They both fail with the same errors from diffing test.rng. When I run the tests locally on both branches, I see the same errors. So I think those branches are broken. Cheers, Martin On 2020-02-12 8:24 a.m., Syd Bauman wrote:
I think it is something local, since at first blush it seems to work in the docker container. But if someone could kick off Stylesheets-dev-against-TEIP5-dev anyway, that would be good. (Or will it kick off automatically when TEIP5-dev is done?)
I don't get this. I just tried a local `make test` in Stylesheets (setting the DEFAULTSOURCE to my local 'dev' copy of p5subst), and it failed because of an expected results error. But I personally copied over the expected results file in question yesterday or the day before.
The diff failure is:
result: ! defined in <ref target="http://www.tei-c.org/release/doc/tei-p5-doc/en/html/BIB.html#XSLT2">6.2. </ref>.</p>
exptected result: ! defined in <ptr target="#XSLT2"/>.</p>
and 1/2 dozen others exactly like that. I.e., w/ full URI instead of just bare name identifier. What gives? Has anyone changed anything that should have anything to do with this?
Is it something on my local system? (That's a rhetorical question.) Could one of the CI maintainers kick off the Stylesheets-dev-against-TEIP5-dev job?
Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
-- ------------------------------------------ Martin Holmes UVic Humanities Computing and Media Centre
Those errors look very much like the Stylesheets are building against the wrong release of the Guidelines. (The difference you see is the change in the regexp used for teidata.word -- "(\p{L}|\p{N}|\p{P}|\p{S})+", which is in test.rng, is the old regexp, whereas "[^\p{C}\p{Z}]+", which you see in expected_results/test.rng, is the new one, present in both dev and release-3.7.0 branches AFAIK.)
Yeah, Syd ‚confessed‘ during last Council call that he already prepared the Stylesheets-dev branch to validate against TEI-dev. So, I’m currently preparing my Jenkins to first create a proper TEI-release-3.7.0 artefact by building against Stylesheets-dev-against-TEIP5-dev. Then (I hope) the Stylesheets-release-7.49.0 job will be able to build against the fresh TEIP5-release-3.7.0 lastSuccessful. Cheers Peter
Am 13.02.2020 um 04:25 schrieb Syd Bauman <s.bauman@northeastern.edu>:
Those errors look very much like the Stylesheets are building against the wrong release of the Guidelines. (The difference you see is the change in the regexp used for teidata.word -- "(\p{L}|\p{N}|\p{P}|\p{S})+", which is in test.rng, is the old regexp, whereas "[^\p{C}\p{Z}]+", which you see in expected_results/test.rng, is the new one, present in both dev and release-3.7.0 branches AFAIK.)
_______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
Good morning, Council! That was the crucial piece we weren’t sure of last night after I made the release branches, but it makes total sense now— we needed to build Stylesheets-dev against TEI-dev. (I think we need to add this to TCW 22...) How are things right now? It looks like Stylesheets is still failing...Do we need to update dev branch with changes to the release branch? (I am going to switch over to Slack...) Elisa Sent from my iPhone
On Feb 13, 2020, at 3:38 AM, Peter Stadler <pstadler@mail.uni-paderborn.de> wrote:
Yeah, Syd ‚confessed‘ during last Council call that he already prepared the Stylesheets-dev branch to validate against TEI-dev. So, I’m currently preparing my Jenkins to first create a proper TEI-release-3.7.0 artefact by building against Stylesheets-dev-against-TEIP5-dev. Then (I hope) the Stylesheets-release-7.49.0 job will be able to build against the fresh TEIP5-release-3.7.0 lastSuccessful.
Cheers Peter
Am 13.02.2020 um 04:25 schrieb Syd Bauman <s.bauman@northeastern.edu>:
Those errors look very much like the Stylesheets are building against the wrong release of the Guidelines. (The difference you see is the change in the regexp used for teidata.word -- "(\p{L}|\p{N}|\p{P}|\p{S})+", which is in test.rng, is the old regexp, whereas "[^\p{C}\p{Z}]+", which you see in expected_results/test.rng, is the new one, present in both dev and release-3.7.0 branches AFAIK.)
_______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
_______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
The release branches on the main Jenkins are fine now (and are reset to build against each other): https://jenkins.tei-c.org/view/Release/ Cheers Peter
Am 13.02.2020 um 09:37 schrieb Peter Stadler <pstadler@mail.uni-paderborn.de>:
Yeah, Syd ‚confessed‘ during last Council call that he already prepared the Stylesheets-dev branch to validate against TEI-dev. So, I’m currently preparing my Jenkins to first create a proper TEI-release-3.7.0 artefact by building against Stylesheets-dev-against-TEIP5-dev. Then (I hope) the Stylesheets-release-7.49.0 job will be able to build against the fresh TEIP5-release-3.7.0 lastSuccessful.
Cheers Peter
Am 13.02.2020 um 04:25 schrieb Syd Bauman <s.bauman@northeastern.edu>:
Those errors look very much like the Stylesheets are building against the wrong release of the Guidelines. (The difference you see is the change in the regexp used for teidata.word -- "(\p{L}|\p{N}|\p{P}|\p{S})+", which is in test.rng, is the old regexp, whereas "[^\p{C}\p{Z}]+", which you see in expected_results/test.rng, is the new one, present in both dev and release-3.7.0 branches AFAIK.)
_______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
Yay! Thank you, Peter! Sent from my iPhone
On Feb 13, 2020, at 5:53 AM, Peter Stadler <pstadler@mail.uni-paderborn.de> wrote:
The release branches on the main Jenkins are fine now (and are reset to build against each other): https://jenkins.tei-c.org/view/Release/
Cheers Peter
Am 13.02.2020 um 09:37 schrieb Peter Stadler <pstadler@mail.uni-paderborn.de>:
Yeah, Syd ‚confessed‘ during last Council call that he already prepared the Stylesheets-dev branch to validate against TEI-dev. So, I’m currently preparing my Jenkins to first create a proper TEI-release-3.7.0 artefact by building against Stylesheets-dev-against-TEIP5-dev. Then (I hope) the Stylesheets-release-7.49.0 job will be able to build against the fresh TEIP5-release-3.7.0 lastSuccessful.
Cheers Peter
Am 13.02.2020 um 04:25 schrieb Syd Bauman <s.bauman@northeastern.edu>:
Those errors look very much like the Stylesheets are building against the wrong release of the Guidelines. (The difference you see is the change in the regexp used for teidata.word -- "(\p{L}|\p{N}|\p{P}|\p{S})+", which is in test.rng, is the old regexp, whereas "[^\p{C}\p{Z}]+", which you see in expected_results/test.rng, is the new one, present in both dev and release-3.7.0 branches AFAIK.)
_______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
_______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
This is really not a workable system at all. The Stylesheets tests should not fail when building against two recent valid versions of TEI. If they're so dependent on the minutiae of a TEI release, then they should be part of the TEI repo. If they're separate, they should successfully test/build against both the current release, the dev branch, and any release branch. A test that fails because TEI changes is not a test of the Stylesheets, it's a test of the TEI. For the moment, I've set my Stylesheets-dev and Stylesheets-release branches to build using the P5-dev result. When those builds are complete, I presume I should set them back to something else, but I'm not sure what the default should be any more. I'm teaching this morning so I won't get much chance to mess around with my Jenkins, unfortunately, so what will probably end up happening is that the Stylesheets-dev and Stylesheets-release branches will (I hope) eventually build successfully, and then the TEI branch builds will kick off using those Stylesheets results. It'll be a long time before all that is completed, though. Cheers, Martin On 2020-02-13 12:37 a.m., Peter Stadler wrote:
Yeah, Syd ‚confessed‘ during last Council call that he already prepared the Stylesheets-dev branch to validate against TEI-dev. So, I’m currently preparing my Jenkins to first create a proper TEI-release-3.7.0 artefact by building against Stylesheets-dev-against-TEIP5-dev. Then (I hope) the Stylesheets-release-7.49.0 job will be able to build against the fresh TEIP5-release-3.7.0 lastSuccessful.
Cheers Peter
Am 13.02.2020 um 04:25 schrieb Syd Bauman <s.bauman@northeastern.edu>:
Those errors look very much like the Stylesheets are building against the wrong release of the Guidelines. (The difference you see is the change in the regexp used for teidata.word -- "(\p{L}|\p{N}|\p{P}|\p{S})+", which is in test.rng, is the old regexp, whereas "[^\p{C}\p{Z}]+", which you see in expected_results/test.rng, is the new one, present in both dev and release-3.7.0 branches AFAIK.)
_______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
This is the circular build process again, then. What version should Stylesheets-dev build against, and what version should Stylesheets-release build against? Cheers, Martin On 2020-02-12 7:25 p.m., Syd Bauman wrote:
Those errors look very much like the Stylesheets are building against the wrong release of the Guidelines. (The difference you see is the change in the regexp used for teidata.word -- "(\p{L}|\p{N}|\p{P}|\p{S})+", which is in test.rng, is the old regexp, whereas "[^\p{C}\p{Z}]+", which you see in expected_results/test.rng, is the new one, present in both dev and release-3.7.0 branches AFAIK.)
Yes indeedy, here we are getting clipped again! :-) I presume, perhaps quite naievely, that dev Stylesheets should test against dev TEI, and dev TEI should build with dev Stylesheets. Furthermore that release-7.49.0 Stylesheets should test against release-3.7.0 TEI, and release-3.7.0 TEI should build with release-7.49.0 Stylesheets. But again, that may be just me forgetting some nuance of this you have explained half a dozen times, but will never sit properly in my brain because it is such a slippery mess. I would definitely like it if and when we get Test2/ working thoroughly and can get it to be less fragile (with respect to small changes in TEI) than Test/ that we just drop Test/ and thus this becomes less of a pain. (Even if the same problem, at least it would not be so annoying.)
This is the circular build process again, then.
What version should Stylesheets-dev build against, and what version should Stylesheets-release build against?
The problem is that you can't build TEI-dev against Stylesheets-dev and Stylesheets-dev against TEI-dev, because that's a circle; one of them has to complete successfully before the other can go ahead. So (as Peter did) we first have to build Stylesheets-dev against something else, then use it to build TEI-dev, then set Stylesheets-dev to build against TEI-dev; and from that point on, as long as nothing breaks, we're OK. But when new release branches are set up, right on the cusp of release day, we instantly have a problem, because they need to build against each other. This happens every time. The release branches should be set up several days ahead of release day, so there's time for all this to settle down. Cheers, Martin On 2020-02-13 7:13 a.m., Syd Bauman wrote:
Yes indeedy, here we are getting clipped again! :-)
I presume, perhaps quite naievely, that dev Stylesheets should test against dev TEI, and dev TEI should build with dev Stylesheets. Furthermore that release-7.49.0 Stylesheets should test against release-3.7.0 TEI, and release-3.7.0 TEI should build with release-7.49.0 Stylesheets. But again, that may be just me forgetting some nuance of this you have explained half a dozen times, but will never sit properly in my brain because it is such a slippery mess.
I would definitely like it if and when we get Test2/ working thoroughly and can get it to be less fragile (with respect to small changes in TEI) than Test/ that we just drop Test/ and thus this becomes less of a pain. (Even if the same problem, at least it would not be so annoying.)
This is the circular build process again, then.
What version should Stylesheets-dev build against, and what version should Stylesheets-release build against?
-- ------------------------------------------ Martin Holmes UVic Humanities Computing and Media Centre
Teilnehmer (4)
-
Elisa Beshero-Bondar -
Martin Holmes -
Peter Stadler -
Syd Bauman