On May 17, 2009, at 10:22 PM, JiHO wrote: Yes... This is quite an interesting area. Krzysztof had made some changes without realizing all the implications of behavior for the different cases. Some needed code has been restored, but the scaling behavior probably won't reappear until the hicolor contents are removed from the build. For testing, you can just find the hicolor directory that contains all the external icons and rename it to something else. Not following sizing from the GTK+ theme settings is one key thing that the external icons are blocking. The good news, however, is that the changes Krzysztof did included one that cleared the last bugs complicating external themes. So now all those files that are under "hicolor" can be dropped in to user directories or system ones as an external theme. So now these can be delivered independent of the official Inkscape release install.

On May 18, 2009, at 1:49 PM, JiHO wrote: Yes, but actually the "Inkscape renders them" solution has a way to address that. The goal is to support some multi-image and/or level-of-detail changes for target sizes. This can include those proposed for SVG 1.2 (I've been told by W3C members that it will be going in) and also the multi-version workflow that jimmac has worked out. We actually would have had that going by now, but instead of being able to support that I have had my time taken away fixing things that were ripped out improperly. Yes. Mainly. But remember that "all other GTK apps" do not have improved SVG rendering in them. Most only have the subset that librsvg has chosen to implement. Do you think you can add a simple chart to the icon page on the wiki? This is really the sort of thing that is easier to sort out there. http://wiki.inkscape.org/wiki/index.php/Icons ***HOWEVER*** the number one point to address with these icon issues is to list them out separately. There are many separate issues that overlap, and not looking at them as separate is the main thing confusing solutions. The times I've been able to find Krzysztof in our chat room it has become clear that this is probably the number one cause of the mistakes he is making. Yes. Those are completely separate things. I was trying to fix that, but then had the icon code changed out from underneath me. The simple solution, though, is only about a single evening's work to implement... as long as the icon code stops being changed from underneath me. You need to list out the details of what system you are seeing that on. Some people have observed no difference, and some have observed broken icons. One rule of performance tuning is that it does not matter how fast you do something, if it's not actually doing the right thing. :-) The caching mental put in was per-run and in-memory only. I later added changes that made the startup render things in the background. That sped startup significantly. Additionally, our work to use the standard XDG directories includes support for their "cache" directory. Hmmm... a standard local directory for apps to "cache" things under???? :-) BINGO!!!! This is JiHO icon issue number 3, not issue number 1 again. (Remember how I said we had separate things overlapping?) The internal rendering (from a single icons.svg file) is 100% compatible with regular theming mechanisms of GTK apps. The bottom line here is that any individual who "themes" other apps can take their exact same steps and theme Inkscape... completely independent of how Inkscape is getting its initial default icons. The reason you see different styles is that Krzysztof did not listen to me, and failed to change to loading "tango_icons.svg" when he changed the other ones to the tango look. This is another place where things start to really break down with external icons (and again, good to toss in a table on the wiki). Hmmm... for the wiki perhaps you can do a table with each feature/ issue down the left side, and "internal", "external" and "notes" across the top. Anyway, there was a script added that was intended to do that. The first problem is that the script is broken, and won't run for many people including myself. This has been pointed out to the person who put in the script, but it has not been fixed. The bigger problem is that the concept itself is wrong. The fixed- sized PNG versions were added as a work-around for systems that did not have librsvg to the latest version. Also.. the source SVG icons themselves were also dumbed down to work for librsvg, and thus lost some Inkscape features. However, only two sizes were done for each icon. So if you happen to be running a theme on a distro that just happens to match those sizes, then you are ok. However... right off the bat our app uses *THREE* gtk sizes, not just two. (however, it happens that Ubuntu sets their default theme to have two of those sizes set to the same value). Additionally, many platforms don't use those exact values, so you get blurry pixmap scaling, missing icons, etc. This also breaks down for those with hi-res displays. Currently I switch between laptops with significantly different DPI (1440x900 compared to 1920x1200 on the same 15.4"), so on one 16x16 might be quite sufficient, but on the other the icons come in too small to be of any real use. It also breaks for any themes tweaked for accessibility such as "Large Print" and similar. No. Basically we can't get fixed bitmaps that will work without adding in thousands of extra images (literally). DOH!!! you're broken. Icons are supposed to be 22x22, not 24x24. :-) Not just bigger screens. Higher DPI screens too. Apple ships products ranging from 94 DPI up through 134 DPI (that's a 43% increase there). And Dell has that one that is 147 DPI, so it's not a rare thing.

On May 23, 2009, at 7:13 PM, Krzysztof Kosiński wrote: Yes, it *IS* compatible with all of those GTK APIs. I have seen it working with my own eyes Did you even bother to test any of your assertions, or are you just making blind guesses??? Please review the code that is in play when icons.svg *is* used. It does use the theme mechanism. Your assertion that there is *no* way to do this is not true. No... those are *not* the only ways to do that. Again, the current icons.svg file code *does* work with icons, falling back, etc. Here too is another key point I need to stress. You keep ignoring this. Please listen this time WE CAN NOT USE LIBRSVG TO GET FULL FEATURES OF SVG!!!! That is *BY DESIGN* of librsvg. They have improved many things recently, but they are intentionally *not* targeting full SVG support. No... Here is another key point. They did not break for *you* but they *DID* break for other people. Please keep in mind that not everyone will be running the latest and greatest versions of all libraries. You *must* keep cross-platform support in mind for any code work you do on Inkscape. One of our key promises is good cross-platform support.

On May 25, 2009, at 5:25 PM, Krzysztof Kosiński wrote: Guess what?!?!?!? THAT IS EXACTLY WHAT WE DO!!!! That is, the code that you declared dead and useless does just such work. So your removal combined with these statements seems to be a clear indication that you did not fully grasp all that was going on in that source file. That is true... However if you read through our source code you'll see that things *are* properly subclassed. Also that is much of the reason for using a factory design pattern. It makes things easy. (Please read up on at least the most common design patterns). Yes, the order does need a bit of tuning. If you look to some of the XDG discussions you'll see that it was something I was asking for assistance with before the other distractions and code changes came up. I believe I also stressed that XDG compatibility with you explicitly. Well... that is your opinion. The opinion of many others is the opposite. Also keep in mind that you have only really been testing with the latest versions of librsvg; older versions have more problems but still are in use. Speed is a non-factor with caching. Memory is a minimal factor due to many things. Icon compatibility is easily achieved with other means that do not dumb things down 100%. Convenient API's is really a red herring, and not a pertinent factor here. Again, on this point you are effectively telling all the artists that they *MUST* work with a constrained, distro-specific set of restrictions. Adoption of such restrictions really should be up to the individual using artist to decide on a job-by-job basis. Here is a point of software engineering. It is usually better to have a simple codepath that supports the needed functionality instead of special-casing various one-off cases, especially if there get to be more of those. Keep It Simple Stupid, or KISS, is how that is often referred to. Finally I'll point out one additional feature proposal that you seem to have missed out on catching. The icon workflow, including the icon preview, is being tweaked to include previews rendered by engines such as librsvg. We don't want to compile this in where we can avoid it (since it does avoid an extra dependency on another external library). But the good news is that we can do this in such a way as to support more than just GNOME. Remember that KDE has it's own very nice SVG rendering. Recent versions of KDE have really jumped ahead in SVG use. So given that the icon workflow is being tweaked to use librsvg through an external hook where possible will simultaneously open it up to support KDE's SVG rendering. Personally I'll be using it with three previews side- by-side: Inkscape internal, librsvg (for GNOME) and KDE. Oh, and in case you didn't know, many of the KDE artists use Inkscape to do their work in.