…ng files (#5116) * fix(iOS): throw FileNotFoundException from openInputStream/openFileInputStream (#1502) The 2015 reporter showed that FileSystemStorage.openInputStream(filename) on iOS would silently return an empty NSFileInputStream for a missing path, because Apple's [NSFileHandle fileHandleForReadingAtPath:] returns nil for non-existent files and NSFileInputStream took the nil without complaint, so callers could not distinguish "missing file" from "legitimately empty file". The dominant path through FileSystemStorage.openInputStream(String) was largely fixed by 2020's commit a97ac44 ("Fixed issue with inconsistent file open behavior on iOS"), which added a fileExists check in openFileInputStream. Two gaps remain: - IOSImplementation.openInputStream(Object) does not check existence when handed a String, so any caller that goes through the generic Object path (e.g. Util.getImplementation().openInputStream(name)) still hits the silent-empty-stream bug. - openFileInputStream throws IOException rather than the more specific FileNotFoundException that JavaSE's FileInputStream throws, so catch (FileNotFoundException) clauses written for cross-platform code silently fall through on iOS. Plug both gaps: openInputStream now does the same fileExists check that openFileInputStream does, and both throw FileNotFoundException instead of plain IOException. Closes #1502. Adds maven/core-unittests/.../FileSystemStorageOpenInputStreamTest.java that drives the public FileSystemStorage API for a guaranteed-missing path and asserts the IOException contract plus the more-specific FileNotFoundException subtype. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(JavaAPI): add java.io.FileNotFoundException + scripts/hellocodenameone test (#1502) The PR previously added FileNotFoundException throws to IOSImplementation but ParparVM's vm/JavaAPI (and Ports/CLDC11) didn't include the class, so the ByteCodeTranslator-emitted Objective-C failed to find java_io_FileNotFoundException.h and the iOS build broke. Add minimal FileNotFoundException stubs to both JavaAPI copies so the translated header exists. Also move the regression test out of maven/core-unittests (where it exercised only JavaSE's FileInputStream contract) to scripts/hellocodenameone/.../tests/, registered in Cn1ssDeviceRunner.DEFAULT_TEST_CLASSES. That harness actually drives the test through the iOS/Android/JavaScript/JavaSE openInputStream paths and is what catches port-level contract regressions on CI. The new test accepts either FileNotFoundException (preferred) or any other IOException (with a log) so it stays green on ports that have not yet narrowed their thrown type, while still failing if any port returns a silent empty stream for a missing path. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * docs: switch FileNotFoundException JavaDoc to /// markdown The CI step .github/scripts/validate-java25-markdown-docs.sh forbids classic /** */ Javadoc markers in CodenameOne/src and Ports/CLDC11/src. The two FileNotFoundException files added for #1502 used the classic form copied from upstream JDK style and were flagged. Convert them to /// markdown comments. No code change. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

on iOS openInputStream for non-existent file do not throw FileNotFoundException #1502

Description

Metadata

Metadata

Assignees

Labels

Type

Fields

Projects

Milestone

Relationships

Development

Issue actions