Repository navigation
Comparing changes
Open a pull request
base repository: qos-ch/logback
base: v_1.6.5
head repository: qos-ch/logback
compare: v_1.6.6
- 16 commits
- 36 files changed
- 5 contributors
Commits on Sep 30, 2026
-
Configuration menu - View commit details
-
Copy full SHA for 8099e42 - Browse repository at this point
Copy the full SHA 8099e42View commit details
Commits on Oct 13, 2026
-
fix: stop the encoder when an OutputStreamAppender is stopped (#1072)
* Stop the encoder when an OutputStreamAppender is stopped OutputStreamAppender.stop() wrote the footer and closed the stream but never stopped its encoder. Encoders that hold resources until stop(), such as logstash-logback-encoder's per-thread JsonGenerator/buffer holders, were therefore never released. Each context reset (e.g. scan="true" reconfiguration) left the previous encoder behind together with one buffer per thread that had logged through it, causing steady heap growth on long-lived applications. The encoder is now stopped after the footer is written and the stream closed. Already stopped encoders are left untouched. Signed-off-by: Samuel Fernandes <[email protected]> * chore: simple unit test Signed-off-by: Samuel Fernandes <[email protected]> --------- Signed-off-by: Samuel Fernandes <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for c33d2cb - Browse repository at this point
Copy the full SHA c33d2cbView commit details -
Configuration menu - View commit details
-
Copy full SHA for 10331ef - Browse repository at this point
Copy the full SHA 10331efView commit details
Commits on Oct 19, 2026
-
Connect SyslogOutputStream's DatagramSocket to the syslog host (#939)
SyslogOutputStream created its DatagramSocket with the no-arg constructor, which binds an ephemeral port on the wildcard address. The appender holds that socket for its whole lifetime, so hardened environments see an unconnected UDP listener on all interfaces (ss -ulnp) for as long as the application runs. This stream only ever sends datagrams to one host:port, so the socket is now connected to that address. A connected UDP socket is reported as connected to its peer instead of listening on the wildcard, while the OS still selects the source address per routing table - remote syslog targets are unaffected. The issue's suggestion of binding to localhost was not followed, as loopback-sourced packets are dropped by remote receivers. SyslogOutputStreamTest asserts the connected state and, on a real local DatagramSocket, that flush() still delivers the datagram. Closes #939
Configuration menu - View commit details
-
Copy full SHA for a4fdeff - Browse repository at this point
Copy the full SHA a4fdeffView commit details -
unit test methods should be public
Signed-off-by: ceki <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for 25d556e - Browse repository at this point
Copy the full SHA 25d556eView commit details -
make connected SyslogOutputStream robust to connect and ICMP errors (#…
…939) Follow-up to 619e102, which connects the DatagramSocket of SyslogOutputStream to the syslog host. - If connect() throws a SocketException, e.g. because the network is not yet available, fall back to an unconnected socket instead of failing the constructor, which previously left SyslogAppenderBase started with a null output stream. SyslogAppenderBase.start() reports the failure as a warning. - On a connected socket, an ICMP port unreachable message caused by an earlier datagram is reported on the next send() as a PortUnreachableException, and the current packet is not sent. flush() now retries once so that no message is lost, as with unconnected sockets. If the retry fails too, SyslogAppenderBase reports a single warning per start() instead of one error per event. - Rename SyslogOutputStream.isInvalidState() to isValidState() (package-private). - SyslogOutputStreamTest: add tests for the connect fallback and for flushing to an unreachable port. The latter is skipped on platforms which do not report ICMP errors on connected datagram sockets. Signed-off-by: ceki <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for 5491685 - Browse repository at this point
Copy the full SHA 5491685View commit details
Commits on Oct 24, 2026
-
Escape regex metacharacters of literal text in FileNamePattern regexes
FileNamePattern.toRegex() and toRegexForFixedDate() copied the literal parts of the file name pattern, and the formatted date, into the regex unchanged. Characters such as '(', '[' or '+' in a directory or file name were therefore interpreted as regex syntax: - RollingFileAppender.start() threw a PatternSyntaxException when the path contained an unbalanced '(', e.g. a Windows user directory such as "C:/Users/Michael(Financ" (#914). - With SizeAndTimeBasedRollingPolicy and a pattern such as "app.%d.(%i).log", the regex did not match existing archives. After a restart within the same period the counter started again at 0, so the next size based rollover renamed the active file onto the existing archive "(0)" and its content was lost. maxHistory and totalSizeCap never removed such archives either (#933). Literal characters with a special meaning in regular expressions are now escaped one by one. Pattern.quote() is not used because callers keep only the part of the regex after the last '/'. The strings returned by toRegex() and toRegexForFixedDate() change accordingly, e.g. "foo-2003.05.20-(\d+).txt" becomes "foo-2003\.05\.20-(\d+)\.txt". Fixes #914 Fixes #933 Signed-off-by: Jialin Chen <[email protected]>Configuration menu - View commit details
-
Copy full SHA for b4b4828 - Browse repository at this point
Copy the full SHA b4b4828View commit details -
Align date pattern to regex conversion with DateTimeFormatter
CachingDateFormatter formats dates with DateTimeFormatter, but the regex conversion and its documentation still assumed SimpleDateFormat. - support the 'L' (stand-alone month) pattern letter in CharSequenceToRegexMapper - add DatePatternToRegexUtil(String, Locale) so that symbols can be taken from a locale other than the default one - FileNamePattern.toRegexForFixedDate(Date) now delegates to the Instant overload - add DatePatternToRegexUtilTest, checking that dates formatted in 16 locales match their regex - check the regex against DateTimeFormatter output in DatePatternToRegexTest Signed-off-by: ceki <[email protected]>Configuration menu - View commit details
-
Copy full SHA for d268774 - Browse repository at this point
Copy the full SHA d268774View commit details -
Escape regex metacharacters in date patterns, harden rolling tests
- escape regex metacharacters in %d{...} date patterns converted to regex; repeated dots now yield \.{n} - RollingFileAppender: warn instead of failing start() when the file/fileNamePattern collision regex is invalid - test archive removal with "(%i)" in fileNamePattern (#933) - MultiThreadedRollingTest: verify output in Java instead of bash, and set a 10 ms check increment so that size based rollover occurs - remove a redundant 100 ms sleep in TimeBasedRollingWithArchiveRemoval_Test - surefire: JUnit module flags for parameterized tests in IDEs; drop parallel/threadCount, which the JUnit Platform provider ignores Signed-off-by: ceki <[email protected]>Configuration menu - View commit details
-
Copy full SHA for 7c36c71 - Browse repository at this point
Copy the full SHA 7c36c71View commit details -
add LinkedHashMap to HardenedObjectInputStream whitelist
Signed-off-by: ceki <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for 5c1f577 - Browse repository at this point
Copy the full SHA 5c1f577View commit details -
LOGBACK-1337 LogbackMDCAdapter should store mdc in LinkedHashMap to …
…predict order of MDC in %X Signed-off-by: Craig P. Motlin <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for 0b95858 - Browse repository at this point
Copy the full SHA 0b95858View commit details -
fix loging of ThrowableProxy constructor taking isCyclic boolean
Signed-off-by: ceki <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for 6217f11 - Browse repository at this point
Copy the full SHA 6217f11View commit details -
reject cyclic throwable proxies in LoggingEventVO.readObject
ThrowableProxyVO.build() only produces trees, but a crafted stream sent to a socket receiver can use serialization back-references to make a cause or suppressed entry point at another proxy, e.g. its own parent. DefaultThrowableRenderer and SyslogAppender then walk the cause chain forever, and the recursive converters overflow the stack. Reject such graphs with InvalidObjectException, next to the existing argument array length check. Signed-off-by: arfa k <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for c7b8e2f - Browse repository at this point
Copy the full SHA c7b8e2fView commit details -
Test round trip of circular exceptions through LoggingEventVO
Check that genuinely circular exceptions, and exceptions reaching the same throwable twice, still deserialize now that cyclic or shared throwable proxies are rejected. Replace reflection in the test with a package-private LoggingEventVO.setThrowableProxyVO(). Signed-off-by: ceki <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for 355b931 - Browse repository at this point
Copy the full SHA 355b931View commit details -
throw ScanException on unmatched '}' during variable substitution, f…
…ixes #836 Previously, Parser.parse() stopped at an unmatched right accolade and silently dropped it along with the rest of the input, e.g. the pattern "%m}some text" was truncated to "%m". The parser now throws a ScanException with the message "Unmatched right accolade '}'". Joran callers already catch ScanException, report an error and keep the original value. Also remove a stray '}' in the input of the oss-fuzz regression test OptionHelperTest.recursionErrorWithNullLiteralPayload, which would now trip the new check before reaching the circular reference detection. Signed-off-by: ceki <[email protected]>
Configuration menu - View commit details
-
Copy full SHA for 7fa8e8e - Browse repository at this point
Copy the full SHA 7fa8e8eView commit details
Commits on Oct 25, 2026
-
Configuration menu - View commit details
-
Copy full SHA for a9ee801 - Browse repository at this point
Copy the full SHA a9ee801View commit details
This comparison is taking too long to generate.
Unfortunately it looks like we can’t render this comparison for you right now. It might be too big, or there might be something weird with your repository.
You can try running this command locally to see the comparison on your machine:
git diff v_1.6.5...v_1.6.6