To me the change to provide pvh for win is just too ugly.
This time, I do not think it is Bill's fault. You insist on 'non-intrusive pch'. As a result, every source file has its own full set of headers and can compile normally. When pch is on, many headers in the source files are shadowed because of --include=pch.h. As far as I can see, the only benefit of this approach is that a few more headers that are not used by *everyone* can be put in pch.h. These headers will not affect non-pch build, and are pre-compiled in pch-mode. However, I do not see much of a problem if the included pch.h has a few more headers either. After all, pch will be turned on by default. The costs are that this is not the usually way people use pch (from what I read about pch) and someone has to read config.m4 to see what is going on; the source files are inflated because many headers can be replaced by pch.h; and of course, this maks lyx less portable (in the case of msvc/pch). (One last problem, this makes scons/pch a little bit more difficult since scons compiles from top source code and has to pass the full path of pch.h to source files.) Unless you can convince me with more benefits of the current approach, I would suggest that lyx follows the mainstream and adds pch.h to source files. Bo