Stream: genomics/committers
Topic: build errors
Patrick Werner (Sep 23 2020 at 12:31):
In one of our recent meetings we talked about the current build is failing. I volunteered to look into it, but can't find the the minutes/meeting we talked about this.
Jamie Jones (Sep 23 2020 at 12:32):
It may have been the co chair meeting
Patrick Werner (Sep 23 2020 at 12:32):
I just tried to build the current master, without a problem
Patrick Werner (Sep 23 2020 at 12:32):
Jamie Jones said:
It may have been the co chair meeting
i thought the same, but couldn't find the location of the co-chair minutes
Jamie Jones (Sep 23 2020 at 12:33):
There is an error that gets logged in the build about medication statement but doesn't list as an actual build error
Jamie Jones (Sep 23 2020 at 12:34):
Build does work currently, Kevin and I did some qa last night and removed ~68 errors
Patrick Werner (Sep 23 2020 at 12:43):
only error i can reproduce (which is only a info msg)
Process Info Validate resource against profile http://hl7.org/fhir/StructureDefinition/MedicationStatement
Jamie Jones (Sep 23 2020 at 12:44):
It shows up in terminal but not as a build error, which makes me think it's just something about how it checks for the new name of the resource first
Patrick Werner (Sep 23 2020 at 12:45):
can't find it in my terminal output. Maybe this was fixed by a recent ig publisher update?
Jamie Jones (Sep 23 2020 at 12:46):
I'm not worked about that one either way :smile:. There are a few indirect valueset url errors that are stumping us though
Jamie Jones (Sep 23 2020 at 12:47):
It will be looking at the right loinc LL url but then append a '|null' and won't resolve
Kevin Power (Sep 23 2020 at 13:03):
I “fixed” the error with MedicationStatement. For some reason an example bundle that had a MedicationStatement caused the problem
Patrick Werner (Sep 23 2020 at 13:07):
So now no MedicationStatement is included anymore?
Kevin Power (Sep 23 2020 at 13:09):
We removed the profile (medicationstatement) from our IG - it didn't seem to provide much value.
And correct, we have no examples including MedicationStatement from FHIR Core. That error we were seeing is most certainly a bug in the IG Publisher.
Patrick Werner (Sep 23 2020 at 13:09):
:+1:
Jamie Jones (Sep 23 2020 at 13:09):
Our profile wasn't constraining any meaningful restrictions on the base resource
Jamie Jones (Sep 23 2020 at 13:10):
It had been constraining it to look essentially like the R3 version of the resource
Patrick Werner (Sep 23 2020 at 13:10):
diagnosticreport-cgexample.xml still contains a MedicationStatement, but doesn't cause problems.
Patrick Werner (Sep 23 2020 at 13:10):
Thanks for fixing my task.
Jamie Jones (Sep 23 2020 at 13:11):
We still saw the message in the build process though
Jamie Jones (Sep 23 2020 at 13:11):
But I think we can ignore it
Kevin Power (Sep 23 2020 at 13:11):
Ah, @Patrick Werner is right - It is a contained resource in that example.
Kevin Power (Sep 23 2020 at 13:12):
I might try restoring one that was causing that weird build error - I have one other idea I wanna try.
Jamie Jones (Sep 23 2020 at 13:12):
It seems the validator does treat contained resources differently
Patrick Werner (Sep 23 2020 at 13:20):
Kevin Power said:
I might try restoring one that was causing that weird build error - I have one other idea I wanna try.
please point me to the problem, want to see wether this is an IG bug
Kevin Power (Sep 23 2020 at 13:21):
Yup, give me 5 mins.
Kevin Power (Sep 23 2020 at 13:24):
One other change we made that perhaps we should discuss - we removed the "extra" slices that were added, to make it more obvious that we could deriveFrom MolecularSequence and "other" Observations. Those were causing MANY errors in qa, so not sure we want to do that.
Patrick Werner (Sep 23 2020 at 13:25):
I think that's fine. There is no technical benefit in these. Should include this into guidance/figures
Kevin Power (Sep 23 2020 at 13:25):
We have some examples that still do this (and in the QA they cause info message like "can't match a slice")
Kevin Power (Sep 23 2020 at 13:28):
Regarding the exception. I happened to noticed this in the "Build Errors" section of the QA file:
The link 'Bundle-HLAB1502-pgx-example.html#MedicationStatement_current-med-example' for "Generated Summary: type: transaction" cannot be resolved (valid targets: 34 targets)
That is when I removed it, and noticed the exception went away as well. I have restored this part of the example now and am doing a commit.
Patrick Werner (Sep 23 2020 at 13:29):
:+1:
Kevin Power (Sep 23 2020 at 13:29):
Commit is complete if you want to pull down and try it @Patrick Werner
Kevin Power (Sep 23 2020 at 13:31):
It is part of a transaction bundle - not sure if that is tripping up something. Or maybe something obvious I am missing.
Kevin Power (Sep 23 2020 at 13:31):
ugg, hang on - I see a mistake.
Kevin Power (Sep 23 2020 at 13:33):
Just noticed I missed removing the meta.profile attribute from this example, and it was still pointing at the old 'medicationstatement' profile we had. It doesn't exist anymore, so honestly would have expected a different error of some sort. I am removing it and rebuilding.
Kevin Power (Sep 23 2020 at 13:34):
OK, never mind - rebuild after removing still generates the exception.
Patrick Werner (Sep 23 2020 at 13:34):
unfortunately Liz and Clem aren't available for the upcoming LOINC quarter
Patrick Werner (Sep 23 2020 at 13:35):
i updated the topics: start with Loinc stuff, then open issues from your agenda @Kevin Power . The ready to vote jira trackers
Kevin Power (Sep 23 2020 at 13:36):
Perhaps we see if Clem/Liz are available for last session of the day, and cover that topic then?
Patrick Werner (Sep 23 2020 at 13:36):
asked liz via email
Kevin Power (Sep 23 2020 at 13:42):
I added two specific JIRAs to the minutes that I think are ready to vote on. @Patrick Werner -- Might need your expertise on how to build a ValueSet from LOINC by attributes in LOINC? See the question in the second to last comment on: https://jira.hl7.org/browse/FHIR-27864
Patrick Werner – Any thoughts on how to most efficiently define a ValueSet based on a LOINC category property that we could use to bind to GenomicsReport(DiagnosticReport).code ?
Patrick Werner (Sep 23 2020 at 13:45):
This topic confuses me. We could create an advanced Filter on Loinc and end up with nothing. As the tooling can't do the expansion on such complex filters.
Easiest would be to include a fixed List into a VS.
Kevin Power (Sep 23 2020 at 13:50):
The concern is that fixed would be hard to maintain
Patrick Werner (Sep 23 2020 at 13:55):
sounds like a discussion we should have in the group
Kevin Power (Sep 23 2020 at 13:55):
Yup - in about 5 minutes or so how about? :smile:
Kevin Power (Sep 23 2020 at 20:20):
@Patrick Werner -- I did notice that "MedicationUsage" actually shows up in the rendered example here:
http://build.fhir.org/ig/HL7/genomics-reporting/Bundle-diagnosticreport-pgxexample.html
Jamie Jones (Sep 23 2020 at 20:22):
(deleted)
Kevin Power (Sep 23 2020 at 20:23):
Patrick Werner (Sep 24 2020 at 13:10):
yes, i saw that. Somehow the IG publisher mirgrates medicationstatemenst to usage. Will debug this on the weekend
Last updated: Apr 12 2022 at 19:14 UTC
