[
https://issues.apache.org/jira/browse/LOG4J2-3358?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17480312#comment-17480312
]
ASF subversion and git services commented on LOG4J2-3358:
---------------------------------------------------------
Commit c062f55318cac4205c7f7606b8e3091f67999516 in logging-log4j2's branch
refs/heads/master from Carter Kozak
[ https://gitbox.apache.org/repos/asf?p=logging-log4j2.git;h=c062f55 ]
LOG4J2-3358: Fix substitutions when programmatic configuration is used (#726)
The problem is that some configuration creation paths fail to
use the Interpolator constructor which loads all plugins. This
produces inconsistent behavior based on the configuration construction
path.
By merging code-paths, we can provide consistent behavior.
> 2.17.1: JsonLayout Context no longer working.
> ---------------------------------------------
>
> Key: LOG4J2-3358
> URL: https://issues.apache.org/jira/browse/LOG4J2-3358
> Project: Log4j 2
> Issue Type: Bug
> Components: Layouts
> Affects Versions: 2.17.1
> Reporter: Markus Duft
> Assignee: Carter Kozak
> Priority: Major
>
> We're creating a RollingFileAppender programmatically using a JsonLayout.
> This uses fields using ${ctx:...} expansion. This stopped working when
> upgrading from 2.17.0 to 2.17.1.
> The code is as follows:
> {code:java}
> /**
> * Creates a rolling file appender that write audit entries to a
> programmatically readable JSON file.
> */
> private RollingFileAppender createJsonAppender(Path logDir) {
> List<KeyValuePair> fields = new ArrayList<>();
>
> fields.add(KeyValuePair.newBuilder().setKey("who").setValue("${ctx:WHO}").build());
>
> fields.add(KeyValuePair.newBuilder().setKey("what").setValue("${ctx:WHAT}").build());
>
> fields.add(KeyValuePair.newBuilder().setKey("method").setValue("${ctx:METHOD}").build());
>
> fields.add(KeyValuePair.newBuilder().setKey("parameters").setValue("${ctx:PARAMETERS}").build());
>
> RollingFileAppender.Builder<?> builder =
> RollingFileAppender.newBuilder();
> builder.setName("auditJsonLogger");
> builder.withFileName(logDir.resolve(LOG_JSON_FILENAME).toString());
>
> builder.withFilePattern(logDir.resolve(LOG_JSON_FILEPATTERN).toString());
> builder.withPolicy(SizeBasedTriggeringPolicy.createPolicy("10M"));
>
> builder.setLayout(JsonLayout.newBuilder().setCompact(true).setEventEol(true).setConfiguration(new
> DefaultConfiguration())
> .setAdditionalFields(fields.toArray(new
> KeyValuePair[0])).build());
> builder.withStrategy(
>
> DefaultRolloverStrategy.newBuilder().withCompressionLevelStr(String.valueOf(Deflater.DEFAULT_COMPRESSION))
>
> .withMax(Integer.toString(LOG_MAX_INDEX)).withFileIndex("min").build());
> return builder.build();
> } {code}
> We also tried to use %X\{...} as well as $${ctx:...} expansions instead to no
> avail. The log file always contains the *literal* string we use in
> .setValue(), i.e. lines like this:
> {noformat}
> {"instant":{"epochSecond":1642756242,"nanoOfSecond":271091900},"thread":"hive/fsck","level":"INFO","message":"Execute","endOfBatch":false,"threadId":65,"threadPriority":5,"who":"${ctx:WHO}","what":"${ctx:WHAT}","method":"${ctx:METHOD}","parameters":"${ctx:PARAMETERS}"}{noformat}
> Reverting to 2.17.0 will correctly give lines like this in the JSON file:
> {noformat}
> {"instant":{"epochSecond":1642756430,"nanoOfSecond":923210700},"thread":"hive/fsck","level":"INFO","message":"Execute","endOfBatch":false,"threadId":62,"threadPriority":5,"who":"marku","what":"ObjectConsistencyCheckOperation","method":"-","parameters":"{dryRun=false,
> roots=[io.bdeploy/demo/client-app/linux:1.0.0,
> io.bdeploy/demo/client-app/linux:2.0.0,
> io.bdeploy/demo/client-app/windows:1.0.0, ...]}"}{noformat}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
