<AWT Dev> 8221405: Fix Windows 32bit awt build
thomas.stuefe at gmail.com
Tue Mar 26 14:41:06 UTC 2019
On Tue, Mar 26, 2019 at 2:36 PM Alexey Ivanov <alexey.ivanov at oracle.com>
> Hi Thomas,
> Looks good, I'm not a reviewer though.
> Could I also ask you to push this changeset to jdk/client  instead of
> A small question: two prototypes are changed in debug_trace.h but only
> one is updated in debug_trace.c. Is it because the functions matching
> the second prototype have already been updated with JNICALL?
Not really sure I understand the question. There is no real 1:1
relationship between my change in .h and .c:
DTRACE_PRINTxx macros call _DTrace_Template() which expands to
DTrace_PrintFunction(). First argument is pointer to DTrace_VPrint(), which
is a function decorated with JNICALL. Hence the build error - the type of
the first argument of DTrace_PrintFunction is DTRACE_PRINT_CALLBACK, which
is a fkt pointer *without* JNICALL.
So one way to fix this was to correct the DTRACE_PRINT_CALLBACK type to be
a fkt pointer pointing to a JNICALL decorated function.
That means that I also needed to fix all functions whose pointers are being
passed around as DTRACE_PRINT_CALLBACK. But there only was one I saw,
I'll push to jdk/client. Will that be transported to jdk/jdk automatically?
>  http://hg.openjdk.java.net/jdk/client/
> On 25/03/2019 13:09, Thomas Stüfe wrote:
> > Hi all,
> > Issue: https://bugs.openjdk.java.net/browse/JDK-8221405
> > cr:
> > On 32bit windows, awt build did bitrot. Dtrace print callbacks need to
> > be declared with __stdcall. No other platform cares.
> > Thanks, Thomas
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the awt-dev