On 2020/11/17 2:16, Daniel Shahaf wrote: > Yasuhito FUTATSUKI wrote on Sat, 14 Nov 2020 14:52 +0900: >> +++ subversion/bindings/swig/INSTALL (working copy) >> @@ -141,7 +141,15 @@ >> - Make sure that Subversion's ./configure script sees your installed SWIG! >> + If you are using the distribution tarball and you want to use the language >> + bindings C source files shipped with it, you might need to pass the >> + --without-swig option to configure script to avoid detecting and checking >> + SWIG on your system. A Makefile generated by configure will prevent >> + building the language bindings if the configure script detect unsuitable >> + version of SWIG. > > I don't dispute the accuracy of this paragraph, but I think this API > isn't autotools-idiomatic. > > Generally, I'd expect --without-foo to short-circuit the probe for foo > and assume foo isn't found; i.e., if my system has an unsuitable > version of foo, I'd the default behaviour (given neither --with-foo nor > --without-foo) and the behaviour given --without-foo to be identical: > namely, behave as though foo isn't available (even if /usr/bin/foo > exists and is perfectly suitable). > > In particular, if my system has an unsuitable version of swig, > I wouldn't expect passing --without-swig to change configure's behaviour.
Probably what is wrong here is that the configure script accepts --with-swig | --without-swig options and checks it in release mode. We never clean SWIG generated language bindings C source files on clean-foo targets in release mode Makefile. extraclean-foo targets do it, but they are only parts of the extraclean target which also removes all release mode stuff. So users never use SWIG in release mode actually. That is, r1876662 is not correct. Cheers, -- Yasuhito FUTATSUKI <futat...@yf.bsclub.org>
