On 22/09/2020 05:11, Paul Edwards wrote: > using JNI. I am getting an error about not being > able to find _imp__JNI_CreateJavaVM at 12 > > I have got conflicting information from the web > as to whether the jvm.lib that is supplied in the > JDK should work or not. Some information says > that it won't work, because the function is > stdcall and mingw32 does not prefix the function > name with an underscore, while MSVC does. Not true. Both mingw32 and MSVC 32-bit prepend one extra underscore to the symbol name, as it appears in C source code, to get the symbol name as it appears in object code, (object files and libraries). Confusion may arise w.r.t. mingw64 and MSVC 64-bit, neither of which prepends the underscore. C++ may also introduce potential for confusion, since the compiler will mangle any symbol name that is not declared to have extern "C" linkage. > Note that the above reference only has one > underscore, instead of the two that show up > when running "nm" on jvm.lib: > > 00000000 I __imp__JNI_CreateJavaVM at 12 > U __IMPORT_DESCRIPTOR_jvm > 00000000 T _JNI_CreateJavaVM at 12 That looks correct; compare the corresponding references for Microsoft's LoadLibraryA(), from our own libkernel32.a: $ nm mingw32/lib/libkernel32.a : 00000000 I __imp__LoadLibraryA at 4 00000000 T _LoadLibraryA at 4 : > I have tried a lot of searches and a lot of > experiments, but nothing successful. I still > can't even answer the basic question of > whether it is meant to work out of the box > or not. I am unaware of any reason why it shouldn't work, and no ... > Or whether I need to construct my own "lib" file. ... you already have an import library, which should work. > Any ideas? Relevant technical info below. Looks like the classical user error of getting the linking order utterly wrong! (It hasn't arisen recently, but used to come up regularly). > C:\paul\devel\jnitest>mingw32-make > mingw32-gcc -o mary.exe \ > -L "C:\Program Files (x86)\Java\jdk1.8.0_221\lib" \ > -ljvm \ This *is* wrong! The linker will read libjvm.a, (having searched for libjvm.dll.a, libjvm.a, and finally jvm.lib, in this order), *before* it has even started building its table of required symbol references, find that this library offers nothing which is required, (because *nothing* is required yet), and dismiss it as irrelevant, without considering it further. > -I "C:\Program Files (x86)\Java\jdk1.8.0_221\include" \ > -I "C:\Program Files (x86)\Java\jdk1.8.0_221\include\win32" \ > mary.c And only now, do you start building a table of required references, *after* jvm.lib has been dismissed as contributing nothing. > c:/mingw/bin/../lib/gcc/mingw32/8.2.0/../../../../mingw32/bin/ld.exe: > C:\Users\xxx\AppData\Local\Temp\ccqlQomh.o:mary.c:(.text+0x3c): > undefined reference to `_imp__JNI_CreateJavaVM at 12' > collect2.exe: error: ld returned 1 exit status > Makefile:7: recipe for target 'mary.exe' failed > mingw32-make: *** [mary.exe] Error 1 I would expect this, because the linking order is *wrong*; the -ljvm reference *must* come *after* all .c or .o files which refer to it. > C:\paul\devel\jnitest>mingw32-gcc --version > mingw32-gcc (MinGW.org GCC-8.2.0-5) 8.2.0 You may wish to consider an upgrade ... we released a mingw32 GCC-9.2.0 in January 2020. > mary.c: > ret = JNI_CreateJavaVM(&jvm, (void**)&env, &args); It would have been useful if you had shown us whatever declaration of JNI_CreateJavaVM() is in scope, when this statement is compiled; we can address your enquiry more effectively, when we are given *all* of the pertinent information ... preferably as an SSCCE (see sscce.org). -- Regards, Keith. Public key available from keys.gnupg.net Key fingerprint: C19E C018 1547 DE50 E1D4 8F53 C0AD 36C6 347E 5A3F -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 833 bytes Desc: OpenPGP digital signature URL: <https://lists.osdn.me/mailman/archives/mingw-users/attachments/20200922/a394b120/attachment.sig>
Fork