On Sun, 2009-08-09 at 12:17 +0200, Mark wrote: > Hi, > > well the 'main' gtk branch should now be more ready for merger as I have > hopefully implemented all of John-Mark's very constructive suggestions > to make the code more robust;
Thanks for the update. I'll attempt to find time tomorrow to look at your changes. > when that is merged, the gtk uploads branch would need reviewing one day > too; This is presumably just the frontend file upload hooks, right? > as for windows branch, the 'clear screen' code that was necessary to > make the background redraw properly, particularly when scrolling, caused > unacceptable flickering; I had to change that particularly for the case > of gifs that caused flickering redraw loops at page load; now I've > simplified the screen clearing to allow the core to simply redraw the > page rather than utterly erasing it first, the background redraw bug has > reappeared; it *looks* as though the core is forgetting background > elements that it has drawn, although the chances are it's more connected > to the communication from the core to the screen > > Be that as it may, I'm hopeful it'll be addressed, possibly I'll need > the assistance of those who may recognise the bug :-) I'm not sure I recognise it from your description. Can you be more specific? > then there are the wayward 'images' that seemed to have appeared > originally when jpegs were being rendered to screen undecoded; so unless > it's connected to libmng - that I'll hope to cross-compile as I'll be > looking into liblcms - then it may be a case for 'those who may > recognise the bug' :-D What is "wayward" about them? > the good news is that windows branch now renders the main image types > including alpha-channel transparency, although the overhead for partial > transparencies of calculating an rgb value - that seems to work at a > reasonable speed in development - may be heavy for older machines, so a > conditional simpler implementation may be necessary If it can be done on 40MHz of ARM, there's precisely no reason why x86 shouldn't be able to cope. J.
