Update facilities schema for new JobManagerType attribute

[Martyn Gigg]

The system tests are failing validation of the Facilities.xml as the new JobManagerType attribute is unknown. We need to update the schema to understand the attribute

[Federico Montesino Pouzols]

add JobManagerType attribute of computeResource in facilities, re #11460

Changeset: ad3a81f7480a879e60f22618cff926dfce1cea2b

[Federico M Pouzols]

Oops, I had totally missed this. It seems that the validation just passes or doesn't really happen on some platforms?

I noticed that the schema defines a few attributes that I'm pretty sure are never used in the code. Maybe we need a tiny design document/discussion about these parameters. I skipped that step as this was just adding one attribute, but in the future more attributes and/or children elements might be added (along the lines of the, currently unused/unknown configFileURL attribute).

[Federico Montesino Pouzols]

add JobManagerType attribute of computeResource in facilities, re #11460

Changeset: 76e57fccdda5c048e1cfcea01667ff529fad891e

[Federico Montesino Pouzols]

This is being verified as ​pull request #501.

[Federico M Pouzols]

Replying to Federico Montesino Pouzols:

add JobManagerType attribute of computeResource in facilities, re #11460

Changeset: ad3a81f7480a879e60f22618cff926dfce1cea2b

This commit had wiped-out the utf8 BOM that the file has at the beginning. Not sure what effects that could have, especially on win7. So I deleted it. Better to leave the file as it is. The following commit should be the good one.

[Martyn Gigg]

The fix here isn't quite right. We've added an attribute not an element type so for this case the update should be a new <xs:attribute name="JobManagerType"/>.

Before we fix this though. I notice that there is already an optional type attribute that can be specified in the Facilities.xsd:31. Was this actually intended for this purpose? If so, then we should just starts using that instead?

[Federico Montesino Pouzols]

My bad, thanks for spotting this. The attribute "type" is never used (as the elements configFileURL, mpirunExecutable, and pythonExecutable) and I had not seen it in the schema. I see two alternatives:

• Use "type"
• Replace "type" with "JobManagerType" or similar

I'd go for the second one, a more specific name. "type" could be confused with other parameters or variants of the schedulers.

I'm also wondering if we have any rules/conventions for capitalization, especially for the first letter, as I see different capitalizations for attributes in different places.

[Martyn Gigg]

I agree that type is a bit generic so let's stick with something more specific here and replace type with jobmanagertype. We seem mostly consistent with XML attributes as all lowercase so let's keep that going.

[Federico Montesino Pouzols]

fix schema, use jobmanagertype (lowercase) attr, re #11460

Changeset: 146e9e99a375df0aa80e582882008cc6f09dac56

[Federico Montesino Pouzols]

use jobmanagertype attr name in parser and tests, re #11460

Changeset: f995b64b6b3819d443e7a24ffcd3093e40a1a936

[Federico Montesino Pouzols]

put BOM back in, re #11460

Changeset: 947685613feb6841a52ec3d41619bde6af10ea28

[Federico Montesino Pouzols]

Good, thanks for your comments. I think this is better now. I'm trying to make sure that everything is fine. But XMl validators apparently don't like our facilities schema (catalog complexType).

[Martyn Gigg]

There is a ValidateFacilitiesFile system test that use some the genxmlif & minixsv python modules to do the validation

[Federico Montesino Pouzols]

Aha, I had to install minixsv which was missing. I think several CI machines are also missing it, and the Validate... system tests are silently skipped.

[Martyn Gigg]

I think you're right. It was installed on the rhel6 machines but not on rhel7

[Federico Montesino Pouzols]

Yes, it seems so. Last night only rhel6 and win7 managed to catch this issue.

[Federico Montesino Pouzols]

I'm now more confident that this is fine. All tests that I could run locally finished happily on win7 and debian, and I also did partial validations of the xml files with other tools.

We got an unrelated unit test failure on osx, a checkout failure on rhel7, and win7 is struggling to catch up with the pull_request build queue...

[Federico Montesino Pouzols]

Jenkins, retest this please

[Martyn Gigg]

Ok. I'll double check it locally and then merge it

[Martyn Gigg]

The windows build failure was an unrelated doctest fail. I'm happy with this now.

[Martyn Gigg]

Merge pull request #501 from mantidproject/11460_update_facilities_schema_JobManagerType_attribute

Add JobManagerType attribute of computeResource in Facilities schema

Full changeset: ​854fdd818ca36904e07bbfa5e65747ba1c6d7b4e

[Nick Draper]

Somehow these slipped through without a resolution. Set to Fixed.

[Stuart Campbell]

This ticket has been transferred to github issue ​12299