RFR: JDK-8013736: [launcher] cleanup code for correctness (original) (raw)

Martin Buchholz martinrb at google.com
Thu May 2 18:16:23 UTC 2013

What would Martin actually do?

#define NULL_CHECK_OR_RETURN(ncor_pointer_to_check,
ncor_failure_return_value)
do {
if ((ncor_pointer_to_check) == NULL) {
JLI_ReportErrorMessage(JNI_ERROR);
return ncor_failure_return_value;
}
} while(0);

On Thu, May 2, 2013 at 10:45 AM, Kumar Srinivasan < kumar.x.srinivasan at oracle.com> wrote:

On 5/2/2013 10:17 AM, Martin Buchholz wrote:

This is global fix creep, but ...

:(

these macros are also found in the hotspot sources. I would rewrite all the macros in the jdk to adopt the blessed style do { ... } while(0) and remove all comments in the jdk of the form /* next token must be ; */ If you want a macro that does nothing at all, you should define it do {} while (0) instead of defining it to the empty string. I am not following, could you be more explicit on how this applies to _-#define NULLCHECK0(e) if ((e) == 0) { +#define NULLCHECKRV(e, rv) if ((e) == 0) { _ _JLIReportErrorMessage(JNIERROR); - return 0; + return rv; _ } -#define NULLCHECK(e) if ((e) == 0) { - JLIReportErrorMessage(JNIERROR); - return; - }+#define NULLCHECK0(e) NULLCHECKRV(e, 0) +#define NULLCHECK(e) NULLCHECKRV(e, )+

I would also be inclined to change == 0 to == NULL Yes will take care of this, as well as Alan suggestion added a check for malloc return. This seems like another occasion to use the weird do { ... } while(0) trick to make the macro behave more like a statement. I might obfuscate the macro parameters to make collisions less likely, e.g. e => NCRVe You want me to rename the macro parameter e to NCRVe ? or something else say ncrve to avoid collision ? Kumar

On Wed, May 1, 2013 at 12:33 PM, Kumar Srinivasan <_ _kumar.x.srinivasan at oracle.com> wrote: Hi, Please review simple fixes for code correctness in the launcher. http://cr.openjdk.java.net/~ksrini/8013736/webrev.0/ Thanks Kumar

More information about the core-libs-dev mailing list