|
From: Allin C. <cot...@wf...> - 2011-01-18 02:00:08
|
On Mon, 17 Jan 2011, [ISO-8859-1] Hans-Bernhard Br wrote: > On 17.01.2011 18:55, Peter Juhasz wrote: > > > The wxt terminal window appears, but without any content, and the > > application doesn't respond anymore. I have to kill it every time, no > > "it magically fixed itself after the fifth time" for me. > > My experience on a real Windows box was somewhat similar, but ultimately > different. The first plot from the first wxt version of gnuplot I ever > ran took _ages_. Yes, it did feel like eternity, but ulimately it got > there. Ever since, it just works. > > So apparently there's something the wxt framework needs to do once to > work properly. Interesting. "My" app gretl ( gretl.sf.net ) calls gnuplot to generate graphs. These are not displayed using the wxt terminal, rather we get gnuplot to generate a PNG file using the pngcairo terminal (which of course shares a lot of code with wxt) and then we display the PNG in a GTK window that's under gretl's control. But we're seeing the same, quite mysterious, issue on Windows: the "first" gnuplot plot takes a long time to appear, where "long" can be anything from 5-10 seconds to well over a minute. It looks like the app is off in the ozone, but eventually the plot appears. I'm not sure exactly what "first" means here -- first plot after booting Windows, first plot after initially installing or updating gretl + gnuplot? -- since that seems to vary. But subsequent plots take a sane amount of time (i.e. barely noticeable delay). My first thought on this was that maybe Windows is conducting some elaborate security audit on the toolchain that's involved (GLib, pango and so on), but that's only a wild guess. I wish I understood what's going on, because it creates a very bad impression for new users. (No such problem on Linux or Mac.) Allin Cottrell |