This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

NLS, attribute errors via cpp during bootstrap


I was going to try looking at Rodney's proposed fixincludes patch this
weekend.  But I can't becuase of bitrot...

 CC=/bin/cc ~/src/egcs/configure --enable-languages=c,c++  
 gmake bootstrap

gmake[3]: Leaving directory `/lfs/home20/robertl/tmp/z/gcc/intl'
test -d po || mkdir po
: --statistics /home/robertl/src/egcs/gcc/po/sv.po -o po/sv.gmo
/bin/sh /home/robertl/src/egcs/gcc/move-if-change po/sv.gmo /home/robertl/src/egcs/gcc/po/sv.gmo
mv: po/sv.gmo: not found
gmake[2]: *** [po/sv.gmo] Error 1

I don't know what $(gmsgfmt) is, but this system doesn't have it.  Thus
po/sv.gmo isn't created (I suppose that's what the -o $@ would do) so the
move-if-change hoses.   So I edit the Makefile to ignore the error and move
on:

/bin/cc -c  -DIN_GCC    -g    -DHAVE_CONFIG_H    -I. -I. -I/home/robertl/src/egcs/gcc -I/home/robertl/src/egcs/gcc/. -I/home/robertl/src/egcs/gcc/config -I/home/robertl/src/egcs/gcc/../include /home/robertl/src/egcs/gcc/c-common.c -o c-common.o
UX:acomp: ERROR: "/home/robertl/src/egcs/gcc/builtin-attrs.def", line 94: Syntax error before or at: _0
UX:acomp: WARNING: "/home/robertl/src/egcs/gcc/builtin-attrs.def", line 94: declaration missing specifiers: assuming "int"
UX:acomp: WARNING: "/home/robertl/src/egcs/gcc/builtin-attrs.def", line 94: declaration missing specifiers: assuming "int"
UX:acomp: ERROR: "/home/robertl/src/egcs/gcc/builtin-attrs.def", line 94: Syntax error before or at: _0
UX:acomp: WARNING: "/home/robertl/src/egcs/gcc/builtin-attrs.def", line 94: declaration missing specifiers: assuming "int"



Looking at the preprocessing output of that compilation, I see:

 ATTR_FORMAT_ARG ,
 
 
# 94
 ATTR_PRINTF_1 _0 , ATTR_FORMAT_PRINTF_1 _0 ,
 ATTR_PRINTF_1 _2 , ATTR_FORMAT_PRINTF_1 _2 ,
 ATTR_PRINTF_2 _0 , ATTR_FORMAT_PRINTF_2 _0 ,
 ATTR_PRINTF_2 _3 , ATTR_FORMAT_PRINTF_2 _3 ,
 ATTR_PRINTF_3 _0 , ATTR_FORMAT_PRINTF_3 _0 ,
 ATTR_PRINTF_3 _4 , ATTR_FORMAT_PRINTF_3 _4 ,
 ATTR_SCANF_1 _0 , ATTR_FORMAT_SCANF_1 _0 ,
 ATTR_SCANF_1 _2 , ATTR_FORMAT_SCANF_1 _2 ,
 ATTR_SCANF_2 _0 , ATTR_FORMAT_SCANF_2 _0 ,
 ATTR_SCANF_2 _3 , ATTR_FORMAT_SCANF_2 _3 ,
 ATTR_STRFTIME_3 _0 , ATTR_FORMAT_STRFTIME_3 _0 ,
 ATTR_STRFMON_3 _4 , ATTR_FORMAT_STRFMON_3 _4 ,
# 107
 
 ATTR_FORMAT_ARG_1 ,


When I pass the same thing through GCC's preprocessor, there is no space
between the "ATTR_PRINTF_1_ and the "_0" so I'm guessing that someone has
a differing idea of how token pasting should be applied.

RJL


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]