RFR: 8292016: Split Windows API error handling from errors passed through the runtime in the JDK [v22]

[David Holmes]

On Mon, 26 Sep 2022 15:47:51 GMT, Julian Waters <jwaters at openjdk.org> wrote:

>> EDIT: Cave and add the ErrorOrigin enum, to differentiate which error type the error reporting functions in libjava will look up. RUNTIME refers to errors passed through the runtime via errno, and SYSTEM is for native errors not visible to the runtime.
>
> Julian Waters has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Whitespace

This seems to be trying to future-proof the reporting code by introducing a duality that is not actually needed at present in some cases. That means the code is untested. Are there planned uses of `RUNTIME` in the very near future?

src/java.base/share/native/libjava/jni_util.h line 339:

> 337:                           char *jniEntryName);
> 338: 
> 339: // RUNTIME - Errors that were passed through the runtime

If "runtime" means "C runtime" (as the system description suggests) then please say that.

src/java.base/share/native/libjava/jni_util.h line 348:

> 346: 
> 347: JNIEXPORT size_t JNICALL
> 348: getLastErrorString(char *buf, size_t len, ErrorOrigin origin);

AFAICS we never pass RUNTIME to `getLastErrorString`

src/java.base/share/native/libjli/java.h line 145:

> 143:  */
> 144: JNIEXPORT void JNICALL
> 145: JLI_ReportErrorMessageSys(ErrorOrigin origin, const char * message, ...);

AFAICS we never pass RUNTIME to `JLI_ReportErrorMessageSys`.

src/java.base/unix/native/libjli/java_md_common.c line 207:

> 205:     va_start(vl, fmt);
> 206:     vfprintf(stderr, fmt, vl);
> 207:     if (errno != 0) {

Is `vfprintf` guaranteed to not set `errno`?

-------------

PR: https://git.openjdk.org/jdk/pull/9870