On 30/11/2020 10:47, Corinna Vinschen wrote:
> On Nov 30 11:25, Corinna Vinschen wrote:
> 
> Also, after applying the patch and autogen-ing, a full build from
> top-level fails with some warnings and a final undefined symbol:

The warnings are expected at this stage.

> make[5]: Entering directory '[...]/x86_64-pc-cygwin/winsup/utils/mingw'
>    CXX      ../bloda.o
>    CXX      ../cygcheck.o
>    CXX      ../dump_setup.o
>    CXX      ../ldh.o
>    CXX      ../path.o
>    CXX      ../cygwin-console-helper.o
>    CXX      ../path_testsuite-path.o
>    CXX      ../strace.o
>    CXX      ../path_testsuite-testsuite.o
> [...]/winsup/utils/mingw/../testsuite.cc:18: warning: "TESTSUITE" redefined
>     18 | #define TESTSUITE

This redefinition should probably be inside #ifndef TESTSUITE/#endif

> <command-line>: note: this is the location of the previous definition
>    CXXLD    cygwin-console-helper.exe
>    CXXLD    ldh.exe
> In file included from [...]/winsup/utils/mingw/../path.cc:263:
> [...]/winsup/utils/mingw/../testsuite.h:22:24: warning: ISO C++ forbids converting a string constant to 'char*' [-Wwrite-strings]
>     22 | #define TESTSUITE_ROOT "X:\\xyzroot"

I'm not sure how to restructure things to avoid this warning.

The '-Wno-error=write-strings' flag is added when building this test to 
avoid this being fatal.

>    CXXLD    path-testsuite.exe
> /usr/lib/gcc/x86_64-w64-mingw32/9.2.1/../../../../x86_64-w64-mingw32/bin/ld: ../path_testsuite-path.o:path.cc:(.rdata$.refptr.max_mount_entry[.refptr.max_mount_entry]+0x0): undefined reference to `max_mount_entry'

This is a bit puzzling.  I don't get this when building locally, but idk 
why since there is only a tentative definition of this variable.

I'm not sure how this being built is changed by automaking to stop it 
working for you (perhaps optimization flags are now being used?)

Perhaps the attached helps, although what is getting stubbed out when 
testing could be clearer.