While trying a cross compile of Python 3.6 I encountered the followingalex@martha:~/src/cpython default☿ hg summaryparent: 101753:31ad7885e2e5  Issue#27225: Fixed a reference leak in type_new when setting __new__ fails.branch: defaultcommit: (clean)update: (current)alex@martha:~/src/cpython default☿ ./configure --build=x86_64-linux --host=arm-linux-gnueabihf --disable-ipv6 ac_cv_file__dev_ptc=no ac_cv_file__dev_ptmx=nochecking for hg... found...creating Makefilealex@martha:~/src/cpython default☿ make...arm-linux-gnueabihf-gcc -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypesParser/acceler.oParser/grammar1.oParser/listnode.oParser/node.oParser/parser.oParser/bitset.oParser/metagrammar.oParser/firstsets.oParser/grammar.oParser/pgen.oObjects/obmalloc.oPython/dynamic_annotations.oPython/mysnprintf.oPython/pyctype.oParser/tokenizer_pgen.oParser/printgrammar.oParser/parsetok_pgen.oParser/pgenmain.o -lpthread -ldl  -lpthread -lutil -oParser/pgenif test "yes" != "yes"; then \Parser/pgen ./Grammar/GrammarInclude/graminit.hPython/graminit.c; \else \cp ./Include/graminit.hInclude/graminit.h; \ficp: './Include/graminit.h' and 'Include/graminit.h' are the same fileMakefile:806: recipe for target 'Include/graminit.h' failedmake: *** [Include/graminit.h] Error 1The attached patch allows me to get past this error. I'm unsure if it's an appropriate fix or just a workaround.

The culprit appears to be the changes introduced forIssue22359 (c36b9ef8ff39 and friends).

Yes, those changes incorrectly assume that the source directory and the build directory are always different (using VPATH). Alex patch LGTM, this other patch overwrites the graminit.[ch] files, not sure if this is better.

I was a little concerned that Alex’s patch used “test A -ef B”, which is not specified by Posix. But it is apparently widely supported (and consider we would only need it for cross compilation). Anyway, I was going to suggest test "$(srcdir)" == "." instead, but maybe Xavier’s solution is slightly better again.I say go with vpath.patch, perhaps with a brief comment explaining that the cp + mv dance avoids copying the file onto itself when doing an in-tree build.

New changeset619f7a2aa40a by Martin Panter in branch '3.5':Issue#27229: Fix in-tree cross-build rule, by Xavier de Gayehttps://hg.python.org/cpython/rev/619f7a2aa40aNew changeset9902230b101f by Martin Panter in branch 'default':Issue#27229: Merge cross-compiling fix from 3.5https://hg.python.org/cpython/rev/9902230b101fNew changesetffed402528c7 by Martin Panter in branch '2.7':Issue#27229: Fix in-tree cross-build rule, by Xavier de Gayehttps://hg.python.org/cpython/rev/ffed402528c7

New changesetec214654708f by Martin Panter in branch '2.7':Issue#27229: Comment in middle of shell command fails on BSD and OS Xhttps://hg.python.org/cpython/rev/ec214654708fNew changesetde5b85f96266 by Martin Panter in branch '3.5':Issue#27229: Comment in middle of shell command fails on BSD and OS Xhttps://hg.python.org/cpython/rev/de5b85f96266New changesetb7a4c076ba40 by Martin Panter in branch 'default':Issue#27229: Merge makefile fix from 3.5https://hg.python.org/cpython/rev/b7a4c076ba40

Thanks Alex & Xavier.