Bruce Dubbs wrote:
> Bruce Dubbs wrote:
>> at least using jhalfs. It seems to build the executable OK, but then
>>
>> makeinfo --split-size=5000000 --split-size=5000000 --no-split -I . -I
>> ../../gcc-4.7.2/gcc/doc \
>> -I ../../gcc-4.7.2/gcc/doc/include -o doc/gccint.info
>> ../../gcc-4.7.2/gcc/doc/gccint.texi
>>
>> gives a lot of warnings in files like:
>
>> make[1]: *** [all] Error 2
>
>> Trying a manual make/make install does not work either. It gives the
>> same errors.
>>
>> I'm looking for something on google, but so far have not found anything
>> recent or anything that looks like a reasonable work around.
>
> I found:
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56258
>
> It doesn't help a lot. It appears that we either edit the Makefile
> (gcc/Makefile.in) to omit the .info file or retain the older texinfo
> package.
>
> Another option may be to add --no-warn and/or --no-validate to the
> MAKEINFOFLAGS, but I haven't tried that.
Those flags do not work. The problem commands are in gcc/Makefile.in:
doc/%.info: %.texi
if [ x$(BUILD_INFO) = xinfo ]; then \
$(MAKEINFO) $(MAKEINFOFLAGS) -I . -I $(gcc_docdir) \
-I $(gcc_docdir)/include -o $@ $<; \
fi
Changing to:
doc/%.info: %.texi
if [ x$(BUILD_INFO) = xinfo ]; then \
touch $@; \
fi
allows the gcc build and install to complete. I didn't recheck the
tests, but they shouldn't be affected.
Another option may be to pass gcc_cv_prog_makeinfo_modern=no to
configure (will create a warning about ).
Still another option is to
sed -i -e 's/BUILD_INFO=info/BUILD_INFO=no/' gcc/configure
Whatever the work around, we will need to do it for both instances of
gcc in Chapter 5 also. Hopefully, this will be fixed in the next
version of gcc.
-- Bruce
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
