Bip

When yacc/bison is not available, make build step fails:

$ make
test -f src/conf.c || /bin/bash ./ylwrap src/conf.y y.tab.c src/conf.c y.tab.h src/conf.h y.output src/conf.output -- yacc -d 
./ylwrap: line 113: yacc: command not found
make: *** [src/conf.c] Error 1

When lex/flex is not available, make build step fails:

$ make
test -f src/lex.c || /bin/bash ./ylwrap src/lex.l .c src/lex.c -- /bin/bash /home/pilou/bip/missing --run flex  
/home/pilou/bip/missing: line 52: flex: command not found
WARNING: `flex' is missing on your system.  You should only need it if
         you modified a `.l' file.  You may need the `Flex' package
         in order for those modifications to take effect.  You can get
         `Flex' from any GNU archive site.
make: *** [src/lex.c] Error 1

Is there a particular reason/problem for preferring a fail during the configure step instead of a fail during the make step ?

According to Autoconf behavior and documentation:

• by default when bison or byacc is not found, the macro AC_PROG_YACC set variable YACC to "yacc": so yacc is considered as always available
• by default when lex or flex are is found, the macro AC_PROG_LEX set variable LEX to ":"

Is inclusion in the bip release tarballs of the files src/conf.h, src/conf.c and src/lex.c generated by bison and flex an acceptable solution?

In fact postponing the error is an intended behavior according to autoconf. Yacc and Lex are optional dependencies, as most software authors publish pre-generated files, and as long as you don't modify them you're not gonna need these tools. Even if we do not publish pre-generated files, i think it's better to keep with this behavior.

Nevertheless, the YACC macro falls back to "yacc" even if it is not present, thus a workaround to use the "missing" mechanism when appropriate is needed.

Applied in changeset 1c3ef1d360c2db1902f5f04415192c123ba4c378.