On Sat, 4 May 2013, Riccardo (Jack) Lucchetti wrote: > On Sat, 4 May 2013, Manny B wrote: > >>> Hold on -- we're econometricians; I'm grateful to Yi-Nung and Sven >>> for testing, but a sample size of 3 cannot yield definitive results. >>> Some bugs are sneaky, only manifesting themselves under certain >>> conditions. If we had 10 tests on Windows XP with only one report of >>> crashing, I would be much happier. >>> >>> It's tempting to say that XP is very old by now (released 11.5 years >>> ago) so why should we bother, but it's (still) currently the second >>> most widely used personal computer operating system after Windows 7. >>> So I take seriously any suggestion that current gretl is unstable on >>> XP; I'd very much like to find out what -- if anything -- is wrong >>> with our gtk stack as packaged for Windows. > > I guess at least 5-10 of my students use XP. If Manny could point out to me a > reliable way to reproduce the bug he's seeing, I could ask for their > cooperation and use them as Guinea pigs.
Manny sent me a photo of the situation, and on that basis this seems to be the recipe for the crash. 1) Start gretl and open a data file. 2) Maximize the main gretl window. 3) Open the model specification dialog (appears on top of the main window). 4) Grab the model dialog by its title bar with the mouse and drag it across the main window. At this point the OS says, "gretlw32.exe encountered a problem". If this is not an effect of some idiosyncratic problem on Manny's system, it seems to me it must have to do with a low-level component in the GTK screen-drawing apparatus, maybe in the libcairo or libpixman DLLs, which have been updated since the 1.9.12 release. Allin
