August 2009 - Inkscape Devel - Inkscape Mailman

Win32 NSIS installer
by Chris Morgan 03 Aug '09

03 Aug '09

Hi all, Today I decided to take a look at the NSIS installer for Win32, and decided to optimise it. It currently uses a macro and function for reading options from the command line. Since then, FileFunc.nsh has had added to it the GetParameters and GetOptions macros, which make the code inside the installer obsolete. I was amused to see that the option-reading code was made by (a different) Chris Morgan in 2004! For your assurance that I hopefully know what I'm doing, I have had lots of NSIS experience, mainly at PortableApps.com, where our launchers and installer are in NSIS. NSIS is an area I can help in, unlike the one I wanted to do more, https://bugs.launchpad.net/inkscape/+bug/398603, which I'd like to fix but can't really. I also optimised readability and similar through the use of LogicLib's logical code structures. I cleaned up the language sections significantly, using a single macro rather than a section for each, containing a macro. In all of them, polng was equal to lng, with the exception of Japanese which had, incorrectly, jp rather than ja for lng. lng was only used for "..\..\inkscape\*.${lng}.txt" anyway - a mask which matches no files for any language. I'd propose that that line be vanquished as it's not used (historical?). In the meantime, I simplified it into one argument for the two. More on languages later. While doing this I found about half a dozen minor bugs, all of which I've fixed. Starting at line 1382 (old), that set of things had the conditions the wrong way around, so it would only print the detail if it was *empty*, rather than if it was something... utterly insignificant, but a bug :D More importantly, all the code around that section was inefficient, I rearranged it to make it more efficient. Also the deleting HKCU\.svg[z] thing was looking for the wrong thing, fixed that. The biggest bug I found was that single-user installations didn't work (at least judging by the code). The shell var context had been formerly set to "all", and the all users section set it to "all" and had a bit of code which wouldn't ever get executed. I got around this the easiest way, which was just to create an unshown section before it to SetShellVarContext current. All in all, I took it down from 1559 lines (47,874 bytes) to 930 lines (34,534 bytes) (60% of the lines, 72% of the filesize), shrunk the installer output by about 25KB (0.07% :P), fixed some bugs and made it more readable in the process. Now for some discussion on what I reckon having looked at it. Languages: I can optimise those still further, just thought I'd send round what I'd got so far. In particular, I want to change it so that no languages are installed by default; this would speed up installation on Windows exponentially - due to the LZMA solid compression used, each language takes an inordinate time to install. I think I can fix that up too, but I'll need to experiment on it a bit. Moving the files first should help in the technique... I'll try it out and report back. However, back on the main topic, I want to change it so English is not shown on the list (it's really pretty meaningless, it's an empty section... and the SectionGroup is "Translations" of the interface, English is pretty much a given), and so that the only language which is installed is that which the installer is running in. I can do that easily enough. Possibly also the MUI LANGDLL should be displayed to let the user choose the initial language. Presets: we've got the SectionIn 1 2 3 bits all over the place, but aren't using the presets at all (any more I guess from commented out code). I think they should be added back, that way you can have all languages in full, the suggested one in optimal and none in minimal. Also I think that quick launch and desktop icons should be on by default. Start menu: currently this is forced on, in a set location, which isn't very nice. We should use the MUI STARTMENU stuff. I also think that a lot of the DetailPrint bits should be cleaned up to either be removed or so they can be more useful. The welcome and finish pages should get their own images, I'll make a proposal for it. Also the header image should get the new logo; likewise I can do that. Right, I think that's enough for now, though I've probably forgotten something I was going to say, and missed a bug or two I fixed. -- Chris Morgan <chris.morganiser@...400...> I don't need a quote in my signature. It's hard enough surviving as it is without having to find a meaningful quote. Will you forgive me? Or don't you read this bit?

6 15

Updated bug hunt tally: 378/500 Bug hunting stalled!!
by Tavmjong Bah 03 Aug '09

03 Aug '09

The current bug hunt tally is now 378/500. Newly fixed bugs: 6 168387 van Lierop Display mode toggle doesn't update radio... 3 396580 JazzyNico non-ASCII characters not displayed... The average number of points per day has dropped to 1 in the past week!! Bug hunting seems to have stalled. Let me know if you think I missed any bugs that should have been credited. Remember, they must be marked Fixed Released or Fixed Committed. Tracked at: http://wiki.inkscape.org/wiki/index.php/Release_Info_for_0.47 Tav

6 8

Boost.Intrusive
by Krzysztof Kosiński 02 Aug '09

02 Aug '09

Hello all In my GSoC project I need to represent nodes in a linked list. The requirements are: 1. Obtaining an iterator from the value should be a constant time operation 2. Ownership semantics should be similar to boost::ptr_list - that is, objects are owned by only one container and deleted when erased, though there are special functions that allow transfer between containers and releasing an object from a container 3. The list must have a 'closed' flag that determines the behavior of 2 functions that retrieve an iterator to the next and previous node: If it's true, they wrap around, and if it's not they behave like regular ++ and -- on iterators. Initially I thought about std::list + std::tr1::unordered_map of pointer to node -> iterator, but this is absurd. Then I created my own NodeList class, which worked well enough, but I feel that implementing all the operations is reinventing the wheel. I could use Boost.Intrusive (specifically boost::intrusive::list) to satisfy 1 and 2 out of the box and 3 with a small amount of work. However, I can't find what version of Boost we currently depend on. Boost.Intrusive is a header only library that was released as part of Boost 1.35 in April 2008 (there are no runtime dependencies, so the only possible issue is whether developers run recent enough systems to have it available). It is available in Debian Lenny, as a backport for Etch, and in Ubuntu Hardy LTS. Is it OK to use it? Regards, Krzysztof Kosiński

5 12

Updated bug hunt tally 420/500 Bug hunt moving again.
by Tavmjong Bah 02 Aug '09

02 Aug '09

The current bug hunt tally is now 420/500. Newly fixed bugs: 3 364224 cilic42 unnecessary tab on layers dialog 9 383244 Hochriener With JessyInk installed in an .svg file, doc prop dialog crash 3 404488 Papouin Axonometric grid spacing lower range is too big 6 386256 JazzyNico Inkscape-0.47pre0 does not build on win32 12 195101 van Lierop stuck in zoom context when typing 'z' while dragging with a tool 9 309856 theAdib crash when doing Python-driven export to a read-only file The average number of points per day has jumped to 10! Let me know if you think I missed any bugs that should have been credited. Remember, they must be marked Fixed Released or Fixed Committed. Tracked at: http://wiki.inkscape.org/wiki/index.php/Release_Info_for_0.47 Tav

2 3

Extracting a saturation map with SVG filters ?
by Ivan Louette 01 Aug '09

01 Aug '09

I can't figure how I could extract a saturation map (as a greyscale image) with SVG filters. My idea is using it to replace any hue of an image with only one like it's done with cyan in the joined png created with Xara Xtreme Linux Edition. I am searching for a long time already but without success. Perhaps not possible. In Xara I only put a cyan rectangle above the image and apply the "Hue" transparency type. I was able to synthezize all Xara's color effects but this one. ivan

2 3

Re: [Inkscape-devel] File Export (Save-As) Tests for v0.47
by ~suv 01 Aug '09

01 Aug '09

On 31/7/09 22:18, Tavmjong Bah wrote: >>> .dfx: Generic AutoCAD export failed to load (see .ai). >> 1) missing helper_extension 'org.inkscape.output.ps' >> (related bug #407215 <https://bugs.launchpad.net/inkscape/+bug/407215>) >> proposed fix: in dxf_output.inx >> s/org.inkscape.output.ps/org.inkscape.print.ps.cairo/ >> >> the 'AutoCAD DXF (*.dxf)'output extension is loaded and creates DXF >> output if pstoedit is installed, though I cannot get pstoedit to create >> DXF files that can be re-imported in Inkscape (on OS X 10.5.7, pstoedit >> 3.45). > > This seems to work but it calls up the PostScript output dialog which I > am not sure is the best. Qcad does read the output dxf OK. Tested on > Fedora 11 with today's svn. AFAIU pstoedit depends on the intermediary .ps format as inkscape output because it does the conversion ps->dxf. Latest version of pstoedit 3.45 is from September 2007. Maybe the helper_extension could to be changed like the one used for the 'print preview' which doesn't prompt for PDF output options either (or does 'print preview' use Postscript output?). Using 'org.inkscape.print.ps' instead of 'org.inkscape.print.ps.cairo' doesn't work although it's defined in src/extension/extension.h. Or there are other cross-platform converters available that can do it in one step: svg->dxf (besides Alvins enhanced dxf_outlines.py extension)? ~suv

2 5