Stream: committers
Topic: Build errors
Grahame Grieve (Mar 18 2016 at 04:49):
committers, note the sequence of build errors right now. This is caused by a subtle and difficult problem in markdown
Grahame Grieve (Mar 18 2016 at 04:51):
in markdown the character < should be escaped if *and only if* it is surrounded by space. It turns out that our library is not compliant, and it is not escaping < even when surrounded by space. It doesn't escape it at all so far as I can see. But we have markdown like this:
Grahame Grieve (Mar 18 2016 at 04:51):
<5ug
Grahame Grieve (Mar 18 2016 at 04:51):
that's illegal in markdown, and, in fact, there's *no legal way* to do that in markdown
Grahame Grieve (Mar 18 2016 at 04:52):
The problem arose because I stop automatically escaping markdown, since it's the wrong thing to do, and a case where it caused problems arose.
Grahame Grieve (Mar 18 2016 at 04:53):
so now I am selectively escaping any < if it is followed by a digit or a = but this is a temporary hack.
Grahame Grieve (Mar 18 2016 at 04:53):
suggestions for real solutions welcome
Grahame Grieve (Mar 18 2016 at 05:06):
no that won't work either. So now, after some fooling around, here's how it works: the tool will pre-escape the content. if you don't want that to happen, you have to prefix your markdown with ```. But either way, you cannot put xml in your markdown. It's not possible, so for as I can see
Paul Knapp (Mar 18 2016 at 11:01):
The build reports the following despite virtuall all examples having the element noted:
[java] INFORMATION: Claim.created: Search Parameter 'Claim.created' had n
o found values in any example. Consider reviewing the path (f:Claim/f:created)
Paul Knapp (Mar 18 2016 at 11:35):
[java] INFORMATION: Claim.organizationreference: Search Parameter 'Claim.
organizationreference' had no found values in any example. Consider reviewing th
e path (f:Claim/f:organizationReference)
Paul Knapp (Mar 18 2016 at 11:46):
[java] INFORMATION: Claim.targetreference: Search Parameter 'Claim.target
reference' had no found values in any example. Consider reviewing the path (f:Cl
aim/f:targetReference)
Bryn Rhodes (Mar 19 2016 at 22:35):
Getting a build error trying to add an element of type Reference(List)
Bryn Rhodes (Mar 19 2016 at 22:36):
[java] ERROR @ Java Compile C:\Users\Bryn\Documents\Src\HL7\FHIR-Dev\implementations\java\org.hl7.fhir.dstu3\src\org\hl7\fhir\dstu3\model\MeasureReport.java:646: error: cannot find symbol
[java] @Child(name = "patients", type = {List_.class}, order=3, min=0, max=1, modifier=false, summary=true)
[java] ^
[java] symbol: class List_
[java] location: class org.hl7.fhir.dstu3.model.MeasureReport.MeasureReportGroupPopulationComponent (src = Publisher)
Bryn Rhodes (Mar 19 2016 at 22:36):
Same definition works fine if I use Reference(Bundle)
Lloyd McKenzie (Mar 26 2016 at 18:59):
I'm vetting the following message:
ERROR @ Questionnaire Resource Questionnaire (FMM=2) cannot have a FMM level >1 if it has warnings (src = ResourceValidator)
However, that's the only error or warning in the build. Has anyone else seen this sort of behavior and if so, any suggestions where I should look for the issue?
Bryn Rhodes (Mar 28 2016 at 21:18):
I see the build is passing, but I'm getting the following error locally:
Exception in thread "main" java.lang.Error: Case mismatch of file C:\Users\Bryn\Documents\Src\HL7\FHIR-Dev\publish\valueset-observation-vitalsignresult.html: found valueset-observation-VitalSignResult.html
[java] at org.hl7.fhir.utilities.CSFile.<init>(CSFile.java:66)
[java] at org.hl7.fhir.utilities.TextFile.stringToFile(TextFile.java:81)
[java] at org.hl7.fhir.tools.publisher.Publisher.generateValueSetPart2(Publisher.java:5394)
[java] at org.hl7.fhir.tools.publisher.Publisher.generateValueSetsPart2(Publisher.java:5353)
Bryn Rhodes (Mar 28 2016 at 21:18):
Any thoughts?
Lloyd McKenzie (Mar 28 2016 at 21:25):
I had that too. I wiped my publish directory and am running again. It seems to be going better. (Or maybe it's just taking longer because it has to create a whole bunch of files now . . .)
Richard Ettema (Mar 28 2016 at 21:45):
I had the same issue. I just deleted all the source/observation files and refreshed via a 'svn update'. All is well on my end now.
Bryn Rhodes (Mar 28 2016 at 22:58):
Okay, I'll try that, thanks!
Lloyd McKenzie (Mar 29 2016 at 00:02):
Hmm. That ran locally. I ran the .bat file to copy stuff but maybe that just did classes and not source files? Manually copying now and I'll recommit
Lloyd McKenzie (Mar 29 2016 at 01:09):
I've broken the build and managed to cost Grahame some local code too. He's now working to rectify and doesn't want me to roll back. So to test any last minute changes, roll back to Grahame's most recent successful build.
Paul Knapp (Mar 31 2016 at 19:52):
Also noticed a pattern where the build cannot find elements from examples in the defined resource if the element name begins with the name of a resource - only a few examples but that appears to be the pattern - if the element is for example observationReporter the error would be 'can't locate and instance of Reporter'.
Grahame Grieve (Mar 31 2016 at 19:54):
can you be more specific about this? I can't visualise where this might be a problem, or where the error might come from
Paul Knapp (Apr 01 2016 at 07:29):
See ExplanationOfBenefit's only example, I have commented out claimResponseReference. We had another element called claimTotal but the element name was change which also resolved the issue. Maybe they were both invalid in some other way, but if you uncomment and build you should get the error.
Grahame Grieve (Apr 07 2016 at 19:59):
@Lloyd McKenzie I've had to make a change to the change you made to Questionnaire. You defined an element 'answered' which is a name clash with 'answer[x]'. The tool isn't supposed to let you do that, but that's why the build is failing
Lloyd McKenzie (Apr 07 2016 at 21:10):
That's not actually a name clash, as "d" is lower-case. However, I can come up with a different name.
Grahame Grieve (Apr 07 2016 at 21:11):
it is a name clash. The rule is that if there's an xxx[n], then no other element can start wtih xxx. That simplifies the logic of working with the definitions greatly
Grahame Grieve (Apr 07 2016 at 21:11):
I renamed it to hasAnswer, but that's not the problem; the failures seem to be that you haven't updated the sdc exmaples...
Lloyd McKenzie (Apr 07 2016 at 21:39):
Well, actually I did update the example. Then I fixed a namespace issue. But unfortunately the namespace issue meant the update didn't run properly. And when diagnosing what happened, I realized there was a problem with my transform. Checking in shortly.
Grahame Grieve (Apr 11 2016 at 05:08):
Brian, shouldn't it be EndPoint not Endpoint?
Brian Postlethwaite (Apr 11 2016 at 23:12):
Happy to update the name of the resource.
(Still needs to pass through FMG too before it can progress to much further)
Grahame Grieve (Aug 02 2016 at 23:33):
I think that the search parameter usage errors are now correct
Grahame Grieve (Aug 02 2016 at 23:33):
e.g . INFORMATION: MedicationAdministration.identifier: Search Parameter 'MedicationAdministration.identifier' had no found values in any example. Consider reviewing the expression (MedicationAdministration.identifier)
Grahame Grieve (Aug 03 2016 at 00:31):
I believe that the warnings about paths that are not found in any example are now correct as of build 9389. Let me know if that isn't the case
Brian Postlethwaite (Aug 03 2016 at 05:39):
Thanks, can't wait to review them!
Brian Postlethwaite (Aug 03 2016 at 05:40):
Much much better!
Thanks Grahame.
Can you have a look at the suppress warnings file for the ones like
Appointment.comment MnM must have confirmed this should not be an Annotation
I couldn't get them suppressed (the ones in the file were approved by MnM already)
Grahame Grieve (Aug 03 2016 at 06:16):
which ones are confimed?
Brian Postlethwaite (Aug 03 2016 at 06:41):
The 6 information messages at the bottom (of suppressed-messages.txt)all still show up. Couldn't work out the format of the line.
(Must be a non-printable char or something in there)
Grahame Grieve (Aug 03 2016 at 07:57):
that particular check doesn't look in suppressed-messages.txt
Brian Postlethwaite (Aug 03 2016 at 12:15):
I'll consider it reported then ;) or do you want me to log it?
Anthony(Tony) Julian (Jan 19 2017 at 22:10):
I synced the build at 03:49 PM and am still getting the build error "Error in StructureDefinition " on all of the resources.
Brian Postlethwaite (Feb 02 2017 at 01:50):
Yes, I did that.
<pre class="xml" fragment="ValueSet">
Is the line that I introduced that causes this failure.
All the other examples on the page have
<pre class="xml" fragment="Coding">
Does this only work for Datatypes, and not resources?
Assuming this is so that the content is validated.
What do I need for resources?
(build\source\datatypes-examples.html)
Lloyd McKenzie (Feb 02 2017 at 01:53):
That's an @Grahame Grieve question . . .
Brian Postlethwaite (Feb 02 2017 at 01:54):
<pre class="xml" fragment="Patient"> works in (extensibility.html)
Brian Postlethwaite (Feb 02 2017 at 01:59):
Looking at it more closely, maybe its doing what its meant to, that example valueset isn't valid!
The identifier property looks nothing like that.
Brian Postlethwaite (Feb 03 2017 at 05:26):
Yes, the error is mine, I'm working on it.
While I'm here, what's the big vomit of output in the middle of the build log that makes it so huge, is that an RDF output of everything? Followed by some weird resolve info logging.
Cooper Thompson (Feb 03 2017 at 18:17):
First commit broken :( working on it
Brian Postlethwaite (Feb 03 2017 at 18:17):
:)
John Moehrke (Feb 03 2017 at 18:19):
welcome to the commit struggle...
Richard Ettema (Feb 03 2017 at 22:06):
@Brian Postlethwaite Yes, that huge output in the middle of the build is from the RDF validation. I've been researching this for the past couple of days and finally tracked it down to a logging issue with the current shex/scala Java libraries. The default logging level without an explicit configuration file is INFO which is where all that output is coming from. I just committed some changes that introduce an explicit logging.properties file to the org.hl7.fhir.tools.core project setting the logging to WARN so all that output is now eliminated. Yay!
Grahame Grieve (Feb 06 2017 at 02:40):
thanks @Richard Ettema. @Brian Postlethwaite I assume that there was no problem in the build tool here
Richard Ettema (Feb 06 2017 at 02:57):
@Grahame Grieve Correct. The build tool is fine. It was a logging issue.
Rick Geimer (Oct 27 2017 at 22:59):
Getting the following error when I run my local build:
ANTLR Tool version 4.5 used for code generation does not match the current runtime version 4.6.
Wondering if others are seeing this too.
Rick Geimer (Oct 30 2017 at 19:27):
Still getting this error locally.after a clean checkout of the repo:
ANTLR Tool version 4.5 used for code generation does not match the current runtime version 4.6.
Any ideas? We seem to have muliple ANTLR versions in tools/java/imports.
Jose Costa Teixeira (Nov 01 2017 at 20:54):
I got a broken build with many errors in resources and examples I did not touch. It worked on my machine.
Jose Costa Teixeira (Nov 01 2017 at 20:56):
am I doing anything wrong? Is there a problem with the build?
Lloyd McKenzie (Nov 01 2017 at 21:28):
@Jose Costa Teixeira Nope, that's not your fault. It's a terminology service issue. I've poked Grahame.
Jose Costa Teixeira (Nov 01 2017 at 21:30):
thanks
it's magically fixed now
Jose Costa Teixeira (Nov 01 2017 at 21:30):
still the log shows that the validation is not going well, but that seems not related to the resources
John Moehrke (Nov 02 2017 at 17:14):
Is there some way we can make the state of the terminology server more visible? Goal to keep all of us committers from frustration when the build failures we have are because the terminology server is down...
Grahame Grieve (Nov 02 2017 at 18:39):
it's never actually down. the problem is more subtle than that
Grahame Grieve (Nov 02 2017 at 18:40):
and I haven't figured out how to debug it yet
Last updated: Apr 12 2022 at 19:14 UTC
