Rob Bearman wrote: >> This will probably mean that Windows users (is that just Michael >> Schmitt?) will suffer from some temporary breakage of the >> Windows port > > I am also building 1.4 for Windows regularly, but only with the Visual > Studio build process, fyi.
Ah yes. I haven't reallly been paying attention to the port of 1.4.x to Windows :) > By the way, will these patches eliminate the > need to apply win32_kludge.diff? They 'eleminate' the need for it in the 1.3.x tree. That's to say, I'm proposing that all the remaining kludges should be added to the tree. That's why lyxserver.C, lyx_cb.C and ispell.C have #ifdef _WIN32 blocks that effectively disable functionality available to the *nix user. Specifically these won't work: * communication with a running LyX by outside programs through the named pipes interface; * spell checking using an external ispell process; * auto-saving of your document every X minutes without blocking your editing session. This last is an ideal candidate for a separate thread and there's no real reason why it shouldn't make its way into the 1.4.x port to Windows. However, I don't really regard it as high priority either. I anticipate that the port of 1.4.x to Windows will be as good as that of 1.3.x but no better. Indeed, I hope that it contains pretty much the same code. Specifically, I propose to rip out the re-implementation of the child process code (SIGCHLD signal handler) that I did last year and insert the 1.3.x code in its place. The current implementation has unsolved 'issues' on nix and just won't work at all on Windows. Conversely, the 1.3.x implementation (explicit polling) is guaranteed to work (and work well) on both platforms. -- Angus
