Hi Yasumasa,
On 17/06/2021 11:28 am, Yasumasa Suenaga wrote:
Hi,
I can build current HEAD of upstream (02c9bf087e5) successfully both
fastdebug and release build.
I use VS 2019 (16.10.1) on WSL 1 Ubuntu 20.04 .
We use 16.9.3 with no problem, but it does seem that the prototype is
missing the static storage class modifier to match the definition
(otherwise it is assumed externa).
I filed:
https://bugs.openjdk.java.net/browse/JDK-8268927
and am testing a trivial fix to add 'static' which will hopefully fix
Vitaly's problem.
Thanks,
David
Thanks,
Yasumasa
On 2021/06/17 10:04, David Holmes wrote:
Hi Vitaly,
On 17/06/2021 9:47 am, Vitaly Provodin wrote:
Hi all,
Building OpenJDK on Windows I am faced with the following error
---------------------------8<---------------------------
os_windows.obj : error LNK2019: unresolved external symbol "int
__cdecl convert_to_unicode(char const *,wchar_t * *)"
(?convert_to_unicode@@YAHPEBDPEAPEA_W@Z) referenced in function
"public: static void __cdecl os::set_native_thread_name(char const
*)" (?set_native_thread_name@os@@SAXPEBD@Z)
That is strange. convert_to_unicode is a static function in
os_windows.cpp so there is no reason for the linker to have any issue
as far as I can see - and we have not seen any build issues locally. ???
I wonder if the forward declaration / prototype also needs to state
static? can you try this change:
diff --git a/src/hotspot/os/windows/os_windows.cpp
b/src/hotspot/os/windows/os_windows.cpp
index 6e996b11993..affe8a10265 100644
--- a/src/hotspot/os/windows/os_windows.cpp
+++ b/src/hotspot/os/windows/os_windows.cpp
@@ -892,7 +892,7 @@ static SetThreadDescriptionFnPtr
_SetThreadDescription = NULL;
DEBUG_ONLY(static GetThreadDescriptionFnPtr _GetThreadDescription =
NULL;)
// forward decl.
-errno_t convert_to_unicode(char const* char_path, LPWSTR* unicode_path);
+static errno_t convert_to_unicode(char const* char_path, LPWSTR*
unicode_path);
void os::set_native_thread_name(const char *name) {
Otherwise the simple fix would be to move the definition of
convert_to_unicode to be ahead of set_native_thread_name() and get rid
of the prototype.
But again I have no idea why we would not have seen this build problem
locally. I will file a bug.
Thanks,
David
-----
c:\buildagent\work\d0555747f6bd5c6\build\windows-x86_64-server-release\support\modules_libs\java.base\server\jvm.dll
: fatal error LNK1120: 1 unresolved externals
make[3]: *** [lib/CompileJvm.gmk:144:
/cygdrive/c/buildagent/work/d0555747f6bd5c6/build/windows-x86_64-server-release/support/modules_libs/java.base/server/jvm.dll]
Error 1
make[2]: *** [make/Main.gmk:252: hotspot-server-libs] Error 2
make[2]: *** Waiting for unfinished jobs....
ERROR: Build failed for targets 'clean images test-image' in
configuration 'windows-x86_64-server-release' (exit code 2)
---------------------------8<---------------------------
The issue was integrated with the patch
https://github.com/openjdk/jdk/commit/9f3c7e74ff00a7550742ed8b9d6671c2d4bb6041
that fixes https://bugs.openjdk.java.net/browse/JDK-8238649 Call new
Win32 API SetThreadDescription in os::set_native_thread_name
After reverting this commit the build completes successfully
Note Visual Studio 2019 Developer Command Prompt v16.8.5 is used for
building
Is this issue actual for me only? - could not find any mentions about
it (in maillists, JBS)
Could you please advice how it can be resolved?
Thanks in advance,
Vitaly
