The dev branch now has all the deprecations valid until 2018-10-01 removed, and its build working ala Travis. However, said build is not working in Jenkins (on either jenkins or jenkins2 -- for some reason the former does not have parsed console output, but it is the same error), and is working on my local system, after a fashion. The error on Jenkins, however, is completely mysterious. It has *nothing* to do with anything that I am aware of having changed. The ONVDL validation of Exemplars/tei_enrich.odd against Exemplars/tei_enrich.nvdl is failing with error: bad value for attribute "columns" where the value for @columns is "1", and thus is perfectly fine. I can, with some effort, reproduce this error on my local system. All of the @columns attrs in tei_enrich.odd are on elements in the teix: namespace, i.e. within <egXML>. I am pretty sure the only schema tei_enrich.nvdl runs against that namespace is tei_enrich-examples.rng. And lo and behold, when I look at that schema the definition of @columns of <layout> is obviously screwball: | <attribute name="columns"> | <a:documentation xmlns:a="http://relaxng.org/ns/compatibility/annotations/1.0">specifies | the number of columns per page</a:documentation> | <list> | <rng:empty xmlns:rng="http://relaxng.org/ns/structure/1.0"/> | </list> | </attribute> I poked around and found that the tei_enrich.odd file changes the definition of <layout> a teeny bit: | <elementSpec ident="layout" mode="change" module="msdescription"> | <attList> | <attDef ident="columns" mode="change" usage="req"/> | </attList> | </elementSpec> But that specification is in a paragraph of prose that essentially negates it: | <p>Note that if (as in the last example above) no value is given for | the <att>columns</att> attribute, the assumption is that there is a | single column of writing on each page. (And is incorrect, in that the prior example does have a value for layout/@columns.) So I claim that a) This is a Stylesheets issue. Somehow changing @columns to required has nuked its datatype, which is supposed to be ( data.count, data.count? ) b) The Enrich ODD has other problems, too I'm too tired now to do much more on this tonight.
Just another note on the ci services: Travis (for the Guidelines repo) is only running the tests, i.e. a `make clean validate test`. Jenkins is running three jobs, a) the tests (same as travis), b) TEIP5-Documentation (which I think we could drop), and c) the complete build Only the last one is failing, hence the difference between travis and Jenkins. Peter PS: Yes, log parsing seems to be broken on Martin’s Jenkins, at present
Am 05.10.2018 um 05:57 schrieb Syd Bauman <s.bauman@northeastern.edu>:
The dev branch now has all the deprecations valid until 2018-10-01 removed, and its build working ala Travis. However, said build is not working in Jenkins (on either jenkins or jenkins2 -- for some reason the former does not have parsed console output, but it is the same error), and is working on my local system, after a fashion.
There seems to be a bug in Jenkins whereby if I define a global log-parsing file, log-parsing fails, but if I specify the same file on a per-project basis, with an explicit system path, it works. It's one of many things that show up as different/broken when moving forward from my old Jenkins (Ubuntu 14.04 to Ubuntu 18.04). Cheers, Martin On 2018-10-05 01:45 AM, Peter Stadler wrote:
Just another note on the ci services: Travis (for the Guidelines repo) is only running the tests, i.e. a `make clean validate test`. Jenkins is running three jobs, a) the tests (same as travis), b) TEIP5-Documentation (which I think we could drop), and c) the complete build
Only the last one is failing, hence the difference between travis and Jenkins.
Peter
PS: Yes, log parsing seems to be broken on Martin’s Jenkins, at present
Am 05.10.2018 um 05:57 schrieb Syd Bauman <s.bauman@northeastern.edu>:
The dev branch now has all the deprecations valid until 2018-10-01 removed, and its build working ala Travis. However, said build is not working in Jenkins (on either jenkins or jenkins2 -- for some reason the former does not have parsed console output, but it is the same error), and is working on my local system, after a fashion.
If we added source="tei:3.4.0“ to the Enrich ODD schemaSpec we’d fix the current build, I believe. I know it’s sort of a hack, but on the other hand we document the „freeze“ of this customisation and keep it stable. What do you think? Peter PS: We’d still need to investigate Syd’s a)
Am 05.10.2018 um 05:57 schrieb Syd Bauman <s.bauman@northeastern.edu>:
The dev branch now has all the deprecations valid until 2018-10-01 removed, and its build working ala Travis. However, said build is not working in Jenkins (on either jenkins or jenkins2 -- for some reason the former does not have parsed console output, but it is the same error), and is working on my local system, after a fashion.
The error on Jenkins, however, is completely mysterious. It has *nothing* to do with anything that I am aware of having changed. The ONVDL validation of Exemplars/tei_enrich.odd against Exemplars/tei_enrich.nvdl is failing with error: bad value for attribute "columns" where the value for @columns is "1", and thus is perfectly fine.
I can, with some effort, reproduce this error on my local system. All of the @columns attrs in tei_enrich.odd are on elements in the teix: namespace, i.e. within <egXML>. I am pretty sure the only schema tei_enrich.nvdl runs against that namespace is tei_enrich-examples.rng. And lo and behold, when I look at that schema the definition of @columns of <layout> is obviously screwball: | <attribute name="columns"> | <a:documentation xmlns:a="http://relaxng.org/ns/compatibility/annotations/1.0">specifies | the number of columns per page</a:documentation> | <list> | <rng:empty xmlns:rng="http://relaxng.org/ns/structure/1.0"/> | </list> | </attribute>
I poked around and found that the tei_enrich.odd file changes the definition of <layout> a teeny bit: | <elementSpec ident="layout" mode="change" module="msdescription"> | <attList> | <attDef ident="columns" mode="change" usage="req"/> | </attList> | </elementSpec> But that specification is in a paragraph of prose that essentially negates it: | <p>Note that if (as in the last example above) no value is given for | the <att>columns</att> attribute, the assumption is that there is a | single column of writing on each page. (And is incorrect, in that the prior example does have a value for layout/@columns.)
So I claim that a) This is a Stylesheets issue. Somehow changing @columns to required has nuked its datatype, which is supposed to be ( data.count, data.count? ) b) The Enrich ODD has other problems, too
I'm too tired now to do much more on this tonight. _______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
Really? (I just checked, yes.) Wow. Two questions immediately jump to mind: 1) What on earth made you think to check such a thing? 1) What is the difference between tei:3.4.0 and tei:current with respect to the @columns attr of <layout>? (The last change to <layout> seems to have been back in July, and to tei_enrich back in January; to <elementSpec> in Nov 2016, and to <attDef> in July). Is a given release of the Guidelines somehow bound to a given release of the Stylesheets? I want the build to work, so I'm all in favor of tying tei_enrich to 3.4.0, or perhaps even something earlier. But then again, I had nothing to do with creating this exemplar. Lou and James, what do you think? Certainly Enrich could use some updating, in any case. The <editionStmt> is clearly false, as is the assertion that it is "a conformant subset of Release 1.1.0", since it extended 1.1.0 by adding <titlePage> to <msContents> (which is no longer necessary).
If we added source="tei:3.4.0“ to the Enrich ODD schemaSpec we’d fix the current build, I believe. I know it’s sort of a hack, but on the other hand we document the „freeze“ of this customisation and keep it stable.
What do you think? Peter
PS: We’d still need to investigate Syd’s a)
I think that, as a general principle, adding a value for @source to indicate the Guidelines release against which any ODD was defined, IS A VERY GOOD IDEA. On 05/10/18 11:48, Peter Stadler wrote:
If we added source="tei:3.4.0“ to the Enrich ODD schemaSpec we’d fix the current build, I believe. I know it’s sort of a hack, but on the other hand we document the „freeze“ of this customisation and keep it stable.
What do you think? Peter
PS: We’d still need to investigate Syd’s a)
Am 05.10.2018 um 05:57 schrieb Syd Bauman <s.bauman@northeastern.edu>:
The dev branch now has all the deprecations valid until 2018-10-01 removed, and its build working ala Travis. However, said build is not working in Jenkins (on either jenkins or jenkins2 -- for some reason the former does not have parsed console output, but it is the same error), and is working on my local system, after a fashion.
The error on Jenkins, however, is completely mysterious. It has *nothing* to do with anything that I am aware of having changed. The ONVDL validation of Exemplars/tei_enrich.odd against Exemplars/tei_enrich.nvdl is failing with error: bad value for attribute "columns" where the value for @columns is "1", and thus is perfectly fine.
I can, with some effort, reproduce this error on my local system. All of the @columns attrs in tei_enrich.odd are on elements in the teix: namespace, i.e. within <egXML>. I am pretty sure the only schema tei_enrich.nvdl runs against that namespace is tei_enrich-examples.rng. And lo and behold, when I look at that schema the definition of @columns of <layout> is obviously screwball: | <attribute name="columns"> | <a:documentation xmlns:a="http://relaxng.org/ns/compatibility/annotations/1.0">specifies | the number of columns per page</a:documentation> | <list> | <rng:empty xmlns:rng="http://relaxng.org/ns/structure/1.0"/> | </list> | </attribute>
I poked around and found that the tei_enrich.odd file changes the definition of <layout> a teeny bit: | <elementSpec ident="layout" mode="change" module="msdescription"> | <attList> | <attDef ident="columns" mode="change" usage="req"/> | </attList> | </elementSpec> But that specification is in a paragraph of prose that essentially negates it: | <p>Note that if (as in the last example above) no value is given for | the <att>columns</att> attribute, the assumption is that there is a | single column of writing on each page. (And is incorrect, in that the prior example does have a value for layout/@columns.)
So I claim that a) This is a Stylesheets issue. Somehow changing @columns to required has nuked its datatype, which is supposed to be ( data.count, data.count? ) b) The Enrich ODD has other problems, too
I'm too tired now to do much more on this tonight. _______________________________________________ Tei-council mailing list Tei-council@lists.tei-c.org http://lists.lists.tei-c.org/mailman/listinfo/tei-council
participants (4)
-
Lou Burnard -
Martin Holmes -
Peter Stadler -
Syd Bauman