On 2018-10-26 03:29, Magnus Ihse Bursie wrote:
I see that you've already pushed this fix. It looks good to me, I was
just wondering about this:
$(info _NT_SYMBOL_PATH=$(_NT_SYMBOL_PATH))
This was not added by your fix, but should we print this always? Looks
like it should be guarded that the log level is info or below.
At this point you are right, it's not that interesting so we shouldn't
always print it.
/Erik
/Magnus
On 2018-10-26 01:27, Erik Joelsson wrote:
When running tests through the new RunTest makefile framework in
Mach5, we are now missing symbol information in hs_err files on
Windows. This is caused by the new run-test-prebuilt target not
properly setting up _NT_SYMBOL_PATH which in turn is caused by the
variable CYGPATH not being set.
This patch sets CYGPATH in RunTestPrebuiltSpec.gmk. I also modified
some cosmetics in the _NT_SYMBOLS_PATH setup to make it easier to
understand, at least for me.
Verfied by running "run-test-prebuilt" with a fastdebug build and
setting JTREG=JAVA_OPTIONS=-XX:ErrorHandlerTest=14. With the fix, the
generated hs_err*.log contains symbols in the stacktrace.
Bug: https://bugs.openjdk.java.net/browse/JDK-8213005
Webrev:
http://cr.openjdk.java.net/~erikj/8213005/webrev.01/make/RunTests.gmk.sdiff.html
/Erik
