Abdelrazak Younes wrote: > Jean-Marc Lasgouttes wrote: >>>>>>> "Georg" == Georg Baum >>>>>>> <[EMAIL PROTECTED]> >>>>>>> writes: >> >> Georg> I is a new bug and a regression WRT 1.4: Now unknown arguments >> Georg> are simply ignored. That should not happen, because that would >> Georg> mean that I can make a typo and don't recognize it. The order >> Georg> of command line evaluation should be the same as in 1.4: >> >> Georg> - Let the frontend parse and remove all arguments that it knows >> Georg> - Parse and remove all known arguments in the LyX singleton - >> Georg> Give an error for all remaining ones.
Actually this is still wrong. The order in 1.4 is: - Parse all known LyX arguments in the LyX singleton and store them - Let the frontend parse and remove all arguments that it knows - Complain for all remaining arguments that are not filenames in the LyX singleton The difference to the procedure above is that it is possible to override frontend options in LyX. >> This is why I wrote that I was afraid of fixing it. The GuiApplication >> object should be instantiated first, but only started if needed. >> >> In particular, if the toolkit has arguments like "-c 3", the "3" will >> be read as a file name... But not if the toolkit eats all known arguments first. Or do I miss something? > Which is bad. Would you accept a change WRT the argument interface? No. I start LyX (and other programs) with filenames many times a day, and always adding -f for LyX would be inconsistent and too much work. > If a > '-f' option was required for documents then all potential problems would > vanish. Why? I could still give wrong arguments that need to be recognized. Georg
