Enrico Forestieri wrote: > On Thu, May 14, 2009 at 09:12:44AM +0200, Abdelrazak Younes wrote: >> Enrico Forestieri wrote: >>> I don't think so. When I tested what was happening when not running in >>> a terminal, I saw that Qt was not spawning the process, so I took care >>> to not use the special devices for redirection in that case. >>> >>> I am a bit concerned about the possible performance impact due to the >>> polling you introduced. I'll also see if that can be alleviated. >>> >> The best way to deal with these output is to use readyReadStandardError >> () and readyReadStandardOutput () signals. I have at hand a >> 'ProcessManager' class that can keep track of all launched processes >> with the option to redirect the outputs to any file we want. For example >> we probably want to distinguish output from LateX and output from the >> image converters (imagemagick...). If you think this is a good idea I >> can try to port it to LyX soonish. > > The point is that when the output goes to a file, it has to be buffered, > otherwise you access frequently the disk. If your disk is networked, you > can understand what a bottleneck you are imposing. Things are different > when the output goes to screen. I think that you should find a way for > fixing the screen output on native Windows (if it really need be; I see > that only the output from configure is irritantly delayed, but otherwise > latex runs are reported in a quite timely way). Heck, one of the reasons > I use cygwin is that of avoiding these unexplicable idionsyncrasies. >
In case of writing to a file using the whole 180 seconds instead of only 500 ms in waitForFinished in will limit the disk access action to only one. Peter
