| Message ID | CAEwic4Y2gnqVfe6POopR7fqEaXg7r2iRfHxuMJYZ5DrW0j+uzw@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
On Feb 9, 2012, at 5:25 PM, Kai Tietz wrote: > Hi, > > this fix ada bootstrap issue for stage 3 for windows native builds > > ChangeLog > > 2012-02-09 Kai Tietz <ktietz@redhat.com> > > PR ada/50942 > * gcc-interface/Makefile.in (LN_S): Add r-option. > > Tested for i686-w64-mingw32. Ok for apply? Puzzling. Aren't these variables only defined on VMS hosts, due to the ifeq just before ? Tristan. > > Regards, > Kai > > Index: gcc/gcc/ada/gcc-interface/Makefile.in > =================================================================== > --- gcc.orig/gcc/ada/gcc-interface/Makefile.in > +++ gcc/gcc/ada/gcc-interface/Makefile.in > @@ -1485,7 +1485,7 @@ ifeq ($(strip $(filter-out alpha64 ia64 > soext = .exe > hyphen = _ > LN = cp -p > -LN_S = cp -p > +LN_S = cp -pr > > .SUFFIXES: .sym
> this fix ada bootstrap issue for stage 3 for windows native builds > > ChangeLog > > 2012-02-09 Kai Tietz <ktietz@redhat.com> > > PR ada/50942 > * gcc-interface/Makefile.in (LN_S): Add r-option. > > Tested for i686-w64-mingw32. Ok for apply? No, this chunk is only relevant for VMS so cannot possibly fix anything on Windows. Can you give the exact error message you are getting? AFAIK, LN_S is/should only be used for files, not directories BTW, so your change would be wrong even if in a Windows section. Arno
2012/2/9 Arnaud Charlet <charlet@adacore.com>: >> this fix ada bootstrap issue for stage 3 for windows native builds >> >> ChangeLog >> >> 2012-02-09 Kai Tietz <ktietz@redhat.com> >> >> PR ada/50942 >> * gcc-interface/Makefile.in (LN_S): Add r-option. >> >> Tested for i686-w64-mingw32. Ok for apply? > > No, this chunk is only relevant for VMS so cannot possibly fix > anything on Windows. > > Can you give the exact error message you are getting? > > AFAIK, LN_S is/should only be used for files, not directories BTW, > so your change would be wrong even if in a Windows section. > > Arno Hmm, I interpret 'ifeq ($(strip $(filter-out alpha64 ia64 dec hp vms% openvms% alphavms%,$(host))),)' as anthing not mentioned here. The bug is as described in the bug-report: Failure on Stage 3 when a make script try to execute following command: cp -p ../.././gcc/ada/rts adainclude Error messgage: cp: omitting directory '../.././gcc/ada/rts' Regards, Kai
On 09/02/2012 18:55, Kai Tietz wrote: > Hmm, I interpret 'ifeq ($(strip $(filter-out alpha64 ia64 dec hp vms% > openvms% alphavms%,$(host))),)' as anthing not mentioned here. "If, after removing alpha64 ia64 dec hp vms% openvms% alphavms% (and stray whitespace) from the host, what remains is equal to nothing" -> "If host consists only of things in alpha64 ia64 dec hp vms% openvms% alphavms%". It's a slightly confusing construct I agree. Watch out for the difference between filter vs. filter-out and ifeq vs. ifneq, and note that that final comma indicates there's an empty string as arg2 of the if condition. cheers, DaveK
> Hmm, I interpret 'ifeq ($(strip $(filter-out alpha64 ia64 dec hp vms% > openvms% alphavms%,$(host))),)' as anthing not mentioned here. That's a misunderstanding as explained by Dave Korn. > The bug is as described in the bug-report: > > Failure on Stage 3 when a make script try to execute following command: > > cp -p ../.././gcc/ada/rts adainclude We should never have such a command emitted by make, so the error is elsewhere/earlier. I suspect some variable didn't get expanded properly. It's possibly a bug in the version/port of make that you are using. Arno
On 9 Feb 2012, at 19:59, Arnaud Charlet wrote: >> The bug is as described in the bug-report: >> >> Failure on Stage 3 when a make script try to execute following >> command: >> >> cp -p ../.././gcc/ada/rts adainclude > > We should never have such a command emitted by make, so the error is > elsewhere/earlier. I suspect some variable didn't get expanded > properly. nevertheless on Darwin (with GNU make 3.81) we get .. rm -rf adalib ln -s ../.././gcc/ada/rts adainclude ln -s ../.././gcc/ada/rts adalib > It's possibly a bug in the version/port of make that you are using. maybe .. but it appears to be wider than just windows. Iain
> nevertheless on Darwin (with GNU make 3.81) we get .. > > rm -rf adalib > ln -s ../.././gcc/ada/rts adainclude > ln -s ../.././gcc/ada/rts adalib Ah, this comes from libada/Makefile.in and is actually "expected", even though it's a bit kludgy IMO. This was introduced by revision 145622 to handle multilibs, I complete forgot about this special handling. In any case, the fix if any belongs either in configure or in libada/Makefile.in, not in gcc-interface. Arno
Index: gcc/gcc/ada/gcc-interface/Makefile.in =================================================================== --- gcc.orig/gcc/ada/gcc-interface/Makefile.in +++ gcc/gcc/ada/gcc-interface/Makefile.in @@ -1485,7 +1485,7 @@ ifeq ($(strip $(filter-out alpha64 ia64 soext = .exe hyphen = _ LN = cp -p -LN_S = cp -p +LN_S = cp -pr .SUFFIXES: .sym