pjfanning opened a new pull request, #125: URL: https://github.com/apache/poi-xmlbeans/pull/125
Fixes https://issues.apache.org/jira/browse/XMLBEANS-676 Ways the `NamespaceContext` thread-local could be left holding memory: - `getCurrent()` created and stored a stack when none existed. `JavaQNameHolder`/`JavaNotationHolderEx` call it outside any push/pop, so the stack was never removed. - `NamespaceContextStack.pop()` set `current` and shrank the stack in two separate steps. An exception between them (e.g. from `Thread.stop`) left them out of step, so the stack never emptied and the thread-local kept a stale `current` (a `TypeStore` holding the whole document's store). - `SchemaParticleImpl` and `SchemaLocalAttributeImpl` called `push` inside the `try`, so a failed push (e.g. OOM) still ran `pop` and removed the caller's frame. On an empty stack `pop` threw `IndexOutOfBoundsException`, which hid the original error. Changes: - `getCurrent()` returns `null` without creating state - `pop()` uses a single `stack.remove(...)`; an unbalanced pop cleans up instead of throwing - `ThreadLocal.remove()` instead of `set(null)` when the stack empties - `push` moved before the `try` in the two schema classes - new `NamespaceContextTest` This can't protect against `Thread.stop` landing between a caller's `push` and its `try`. Callers in that situation should use `ThreadLocalUtil.clearAllThreadLocals()`. 🤖 Generated with [Claude Code](https://claude.com/claude-code) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
