On 2020-04-14 15:13, Erik Joelsson wrote:
The intention is that the new dll should only be required for
TOOLCHAIN_VERSION that define it, just like MSVP_DLL which was
introduced in VS2012.
I will verify again that this is actually true.
My worry is with this piece of code in toolchain_windows.m4:
+ if test "x$VCRUNTIME_1_NAME" != "x"; then
+ if test "x$with_vcruntime_1_dll" != x; then
+ # If given explicitly by user, do not probe. If not present, fail
directly.
+ TOOLCHAIN_CHECK_POSSIBLE_MSVC_DLL($VCRUNTIME_1_NAME,
[$with_vcruntime_1_dll],
+ [--with-vcruntime-1-dll])
+ if test "x$MSVC_DLL" = x; then
+ AC_MSG_ERROR([Could not find a proper $VCRUNTIME_1_NAME as specified
by --with-vcruntime-1-dll])
+ fi
+ VCRUNTIME_1_DLL="$MSVC_DLL"
+ elif test "x$DEVKIT_VCRUNTIME_1_DLL" != x; then
+ TOOLCHAIN_CHECK_POSSIBLE_MSVC_DLL($VCRUNTIME_1_NAME,
[$DEVKIT_VCRUNTIME_1_DLL], [devkit])
+ if test "x$MSVC_DLL" = x; then
+ AC_MSG_ERROR([Could not find a proper $VCRUNTIME_1_NAME as specified
by devkit])
+ fi
+ VCRUNTIME_1_DLL="$MSVC_DLL"
Unless I'm reading this incorrectly, it looks like this will end in an
error if the dll is not found. But maybe I'm just confused.
... ahhh ... right... I'm confused. It says DEVKIT_VCRUNTIME_1_DLL in
the test. So it fails if the devkit is badly configured.
Ok, fine.
The patch LGTM now. (It was always good, but it's first now it looks
that way to me ;-))
/Magnus
/Erik
On 2020-04-14 05:59, Magnus Ihse Bursie wrote:
Maybe I'm missing something here, but will this not break for VS
versions before 2019 that does not have vcruntime140_1.dll?
/Magnus
On 2020-04-09 18:56, Erik Joelsson wrote:
When building with VS2019, there is a new runtime library,
vcruntime140_1.dll, that we need to bundle along with the others.
This patch sets this up all through the build.
Bug: https://bugs.openjdk.java.net/browse/JDK-8242468
Webrev: http://cr.openjdk.java.net/~erikj/8242468/webrev.01/index.html
/Erik
