Jean-Marc Lasgouttes wrote: > Angus> Jean-Marc Lasgouttes wrote: I need this patch to compile the > Angus> 1.4.x tree with mingw. It results in support/globbing.C being > Angus> compiled only if we're building the XForms frontend. >>> Why don't you just move globbing.* to frontends/xforms? > > Angus> Good idea, actually. But the client sub directory stuff is > Angus> still needed. > > Yes, I was about to answer to your previous message, which described > your plans. Basically, I think this should not be done by testing for > mingw, which is against autoconf usage, but by checking the feature > you need. If the server and the client rely on select() being > available, then it is _this_ feature that you should test to decide > whether to compile this stuff. The same holds for ispell, of course.
Ok. So a configure test for HAVE_SELECT that also results in AM_CONDITIONAL(BUILD_CLIENT_SUBDIR, test x$have_select = xtrue) I've come round to the opinion that ISpell.C should have a stub implementation for the case when HAVE_SELECT is not defined. The existing ISpell interface is sufficient to enable the rest of LyX to know that communication with the ispell binary has failed. That just leaves lyxserver and lyxsocket. Again, I think that we need stub implementations in the .C files. -- Angus