- Inkscape Devel - Inkscape Mailman

Re: [Inkscape-devel] Memory Problems
by Richard Kowalczyk 07 May '05

07 May '05

Results of comparing zoomed out with regular zoom: regular: memory percent: 6.6% -> 23.6% buffers/cache: 224 -> 310 zoomed out: memory percent: 5% -> 12.5% buffers/cache: 217 -> 254 The zoomed out numbers are very similar to the numbers for just previewing each file. So it appears that you are correct that the memory problem is in the renderer. On Sat, 2005-05-07 at 11:32 -0300, bulia byak wrote: > Interesting, this means that normal loading is indeed different enough > from previewing, memory-wise. We had assumed it's mostly the same, so > we didn't test multiple file loading except via previews (to my > knowledge). > > It may be that the source of the leaking is the renderer, which may > leak more memory for full-screen display than for generating previews > (even though the same renderer is used in both cases). Can you please > try one more experiment: in all your files, zoom out to minimum zoom > and save them under different names with that zoom. Then, load them > all normally one after the other into the same Inkscape instance. If > the renderer is the culprit, this should have some measurable effect > on memory consumption. > > On 5/7/05, Richard Kowalczyk <richk@...805...> wrote: > > I just tried what you suggested, and it seems to use significantly less > > memory. > > > > I had some different apps open this time, so I can't compare it exactly > > with my previous tests, but here is a rough comparison: > > > > opening all files: > > memory percent: 7% -> 45% > > buffers/cache: 214 -> 435 > > > > opening and closing the previous: > > memory percent: 5.9% -> 27.9% > > buffers/cache: 231 -> 342 > > > > inkview: > > memory percent: 3.5% -> 21.9% > > buffers/cache: 226 -> 327 > > > > inkscape preview window: > > memory percent: 5% -> 13% > > buffers/cache: 337 -> 372 > > > > So just previewing the files seems to use up significantly less memory > > than even inkview. > > > > On Sat, 2005-05-07 at 10:58 -0300, bulia byak wrote: > > > Richard, can you also try this: in the Open dialog, just select these > > > files one by one, so that a preview of each one is shown, and see if > > > there's a memory difference at the end. (Each preview loads and > > > renders the entire document.) I do remember that we fixed, at least > > > partially, the memory leaking when going through previews in this way. > > > So what you describe may be explained by: 1) our not fixing the leaks > > > completely, only partially, or 2) Inkscape using a different and > > > not-yet-fixed load path when loading the document fully than when > > > doing a preview. > > > > > > Mental, can you comment? > > > > > > > > > On 5/5/05, Richard Kowalczyk <richk@...805...> wrote: > > > > I tried doing what you suggested. I went though 26 slides, since it > > > > seemed to be enough to get an indication of the behavior. > > > > > > > > (Just to clarify - I opened the next one (from within the current one), > > > > then closed the current one - if I had closed the current one first, the > > > > memory would be cleared) > > > > > > > > The memory used seems to be roughly equivalent between inkscape and > > > > inkview, although inkview does use less. Either way though, the memory > > > > keeps going up as I go through more slides. > > > > > > > > I attached three files to this email - the result of doing the command: > > > > > > > > ps aux | grep -e "inkview" >> memory.txt; free -m >> memory.txt > > > > > > > > for three cases: > > > > > > > > memory_inkview.txt: after I clicked the right arrow to advance each > > > > slide, I ran the command. > > > > > > > > memory_one_per_time.txt: I did what you suggested, and opened each file > > > > in inkscape, then closed the old one, running the command for each new > > > > file. > > > > > > > > memory_all_open.txt: I opened each file in inkscape, but didn't close > > > > the previous file, so by the end, I had 26 files open at once. I ran > > > > the command after each file was opened. > > > > > > > > Thanks for your help, > > > > Rich. > > > > > > > > On Thu, 2005-05-05 at 19:38 -0300, bulia byak wrote: > > > > > On 5/5/05, Richard Kowalczyk <richk@...805...> wrote: > > > > > > I will be glad to do anything else that anyone thinks will help explain > > > > > > the memory usage. > > > > > > > > > > Can you try loading these same files in turn (i.e. closing the > > > > > previous one when loading the next) into Inkscape instead of Inkview > > > > > and see if there is a noticeable difference? > > > > > > > > > -- > > > > Richard Kowalczyk <richk@...805...> > > > > > > > > > > > > > > > > > > > > -- > > Richard Kowalczyk <richk@...805...> > > > > ------------------------------------------------------- > > This SF.Net email is sponsored by: NEC IT Guy Games. > > Get your fingers limbered up and give it your best shot. 4 great events, 4 > > opportunities to win big! Highest score wins.NEC IT Guy Games. Play to > > win an NEC 61 plasma display. Visit http://www.necitguy.com/?r=20 > > _______________________________________________ > > Inkscape-devel mailing list > > Inkscape-devel(a)lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/inkscape-devel > > > > -- Richard Kowalczyk <richk@...805...>

2 1

Re: [Inkscape-devel] Memory Problems
by Richard Kowalczyk 07 May '05

07 May '05

I just tried what you suggested, and it seems to use significantly less memory. I had some different apps open this time, so I can't compare it exactly with my previous tests, but here is a rough comparison: opening all files: memory percent: 7% -> 45% buffers/cache: 214 -> 435 opening and closing the previous: memory percent: 5.9% -> 27.9% buffers/cache: 231 -> 342 inkview: memory percent: 3.5% -> 21.9% buffers/cache: 226 -> 327 inkscape preview window: memory percent: 5% -> 13% buffers/cache: 337 -> 372 So just previewing the files seems to use up significantly less memory than even inkview. On Sat, 2005-05-07 at 10:58 -0300, bulia byak wrote: > Richard, can you also try this: in the Open dialog, just select these > files one by one, so that a preview of each one is shown, and see if > there's a memory difference at the end. (Each preview loads and > renders the entire document.) I do remember that we fixed, at least > partially, the memory leaking when going through previews in this way. > So what you describe may be explained by: 1) our not fixing the leaks > completely, only partially, or 2) Inkscape using a different and > not-yet-fixed load path when loading the document fully than when > doing a preview. > > Mental, can you comment? > > > On 5/5/05, Richard Kowalczyk <richk@...805...> wrote: > > I tried doing what you suggested. I went though 26 slides, since it > > seemed to be enough to get an indication of the behavior. > > > > (Just to clarify - I opened the next one (from within the current one), > > then closed the current one - if I had closed the current one first, the > > memory would be cleared) > > > > The memory used seems to be roughly equivalent between inkscape and > > inkview, although inkview does use less. Either way though, the memory > > keeps going up as I go through more slides. > > > > I attached three files to this email - the result of doing the command: > > > > ps aux | grep -e "inkview" >> memory.txt; free -m >> memory.txt > > > > for three cases: > > > > memory_inkview.txt: after I clicked the right arrow to advance each > > slide, I ran the command. > > > > memory_one_per_time.txt: I did what you suggested, and opened each file > > in inkscape, then closed the old one, running the command for each new > > file. > > > > memory_all_open.txt: I opened each file in inkscape, but didn't close > > the previous file, so by the end, I had 26 files open at once. I ran > > the command after each file was opened. > > > > Thanks for your help, > > Rich. > > > > On Thu, 2005-05-05 at 19:38 -0300, bulia byak wrote: > > > On 5/5/05, Richard Kowalczyk <richk@...805...> wrote: > > > > I will be glad to do anything else that anyone thinks will help explain > > > > the memory usage. > > > > > > Can you try loading these same files in turn (i.e. closing the > > > previous one when loading the next) into Inkscape instead of Inkview > > > and see if there is a noticeable difference? > > > > > -- > > Richard Kowalczyk <richk@...805...> > > > > > > > > -- Richard Kowalczyk <richk@...805...>

1 0

Re: [Inkscape-user] eps error (bad encoding)
by bulia byak 07 May '05

07 May '05

On 5/2/05, Aewyn <aewyn@...326...> wrote: > I just try to print from inkscape; as I know the only way to > produce a reasonable sized document is save as .eps > > Is is ok, but all my accent characters (őúéáűó) goes to its > unicode formats, so the result is totally unusable. Convert text to path - this is perhaps the most reliable solution. Other than that, I don't know what encoding the text in PS is supposed to be. I doubt it's Unicode. Latin-1? What about non-Western-European languages? Can any experts comment? -- bulia byak Inkscape. Draw Freely. http://www.inkscape.org

3 2

listener fixes
by MenTaLguY 07 May '05

07 May '05

Ok ... I think I fixed all the bugs in CompositeNodeObserver, so I've comitted my patch. I've not gone through and verified that it fixes all the layer etc. bugs yet, but if it doesn't it should at least make the remaining work of fixing them much easier. The commit ensures that two of my "five invariants" are now observed: * a listener should not receive events that occurred before its registration * a listener should not receive any events at all once it is deregistered Hmm, and actually also that "fifth and a half" invariant: it should be safe for a listener to unregister itself. The patch also had the effect of disabling the 'is_interactive' flag for attribute changes. However, the work I'd done since it had been introduced made a (much cleaner!) replacement very easy. [ I've not actually entirely purged the is_interactive flag yet, since that will require changing a lot of files, but I will do so once we're satisfied that these changes are working out. ] I've been testing most of the night, but please test yourself and see if you turn up anything. I'm still worried that some code somewhere was written that inadvertantly depended on a violation of the first invariant above. -mental

1 0

Extensions Update?
by unknown＠example.com 07 May '05

07 May '05

I just thought I would check in and see if there is anything new going on in the extension realm. (Ted, have you found any time to start working on the interpreter property and absolute paths?) I have recently created an effect that draws microtonal and multiscale stringed instrument fretboards. Sounds very useful doesn't it :) I have also started working on an effect to do isometric path extrusions. I think may have taken this idea as far as I am able. Anyone is welcome to pick it up and improve it if they wish. Besides Multiline text boxes, here are a few other extension autogui widgets I would find useful. Any comments? - Enumerated choices (drop down list box) - File selector - Mutually exclusive parameter groups (tabs) like: <tabset> <tab> <param name="length"></param> </tab> <tab> <param name="firstlength"></param> <param name="lastlength"></param> </tab> </tabset> And lastly Inkscape crashes on startup if an *.inx file contains a string param with an empty default value like this: <param name="crasher" type="string"></param> Starting program: /usr/local/bin/inkscape [Thread debugging using libthread_db enabled] [New Thread -1223995264 (LWP 7496)] Program received signal SIGSEGV, Segmentation fault. [Switching to Thread -1223995264 (LWP 7496)] 0x0836725f in ParamString (this=0x85ba1e0, name=0x0, guitext=0x0, ext=0x0, xml=0x862480c) at repr.h:175 175 repr.h: No such file or directory. in repr.h (gdb) bt #0 0x0836725f in ParamString (this=0x85ba1e0, name=0x0, guitext=0x0, ext=0x0, xml=0x862480c) at repr.h:175 #1 0x08366b30 in Inkscape::Extension::Parameter::make (in_repr=0x862480c, in_ext=0x85ba070) at extension/parameter.cpp:238 #2 0x083627d4 in Extension (this=0x85ba070, in_repr=0x0, in_imp=0x85a722e) at extension/extension.cpp:92 #3 0x0836c66b in Effect (this=0x85ba070, in_repr=0x0, in_imp=0x0) at extension/effect.cpp:26 #4 0x0836a258 in build_from_reprdoc (doc=0x84a859c, in_imp=0x0) at extension/system.cpp:418 #5 0x0836a3f0 in Inkscape::Extension::build_from_file ( filename=0x859f980 "/home/aaron/.inkscape/extensions/ffss.inx") at extension/system.cpp:447 #6 0x08365d22 in build_module_from_dir (dirname=0x85a86e0 "/home/aaron/.inkscape/extensions") at extension/init.cpp:207 #7 0x08365b81 in Inkscape::Extension::init () at stl_vector.h:501 #8 0x08122acc in inkscape_application_init (argv0=0x0, use_gui=0) at inkscape.cpp:580 #9 0x08117b6a in sp_main_gui (argc=1, argv=0xbffff9c4) at main.cpp:517 #10 0x081ec4fb in Inkscape::NSApplication::Application::run (this=0x0) at application/application.cpp:132 #11 0x08117703 in main (argc=1, argv=0xbffff9c4) at main.cpp:414 Aaron Spike

1 0

Re: [Inkscape-devel] Memory Problems
by Richard Kowalczyk 06 May '05

06 May '05

I tried doing what you suggested. I went though 26 slides, since it seemed to be enough to get an indication of the behavior. (Just to clarify - I opened the next one (from within the current one), then closed the current one - if I had closed the current one first, the memory would be cleared) The memory used seems to be roughly equivalent between inkscape and inkview, although inkview does use less. Either way though, the memory keeps going up as I go through more slides. I attached three files to this email - the result of doing the command: ps aux | grep -e "inkview" >> memory.txt; free -m >> memory.txt for three cases: memory_inkview.txt: after I clicked the right arrow to advance each slide, I ran the command. memory_one_per_time.txt: I did what you suggested, and opened each file in inkscape, then closed the old one, running the command for each new file. memory_all_open.txt: I opened each file in inkscape, but didn't close the previous file, so by the end, I had 26 files open at once. I ran the command after each file was opened. Thanks for your help, Rich. On Thu, 2005-05-05 at 19:38 -0300, bulia byak wrote: > On 5/5/05, Richard Kowalczyk <richk@...805...> wrote: > > I will be glad to do anything else that anyone thinks will help explain > > the memory usage. > > Can you try loading these same files in turn (i.e. closing the > previous one when loading the next) into Inkscape instead of Inkview > and see if there is a noticeable difference? > -- Richard Kowalczyk <richk@...805...>

1 0

new debugging aid
by MenTaLguY 05 May '05

05 May '05

I've introduced a logging facility; the goodies are in the Inkscape::Debug namespace. Set the INKSCAPE_DEBUG_LOG environment variable to a filename, and Inkscape will write a structured XML outline of its activities there. You can take advantage of this when debugging your own code. For example, if you want to highlight the changes that happen during a function, you add this to the top: Debug::EventTracker<Debug::SimpleEvent> tracker("my-func"); (if the function isn't inside the ::Inkscape namespace, you'll need to fully-qualify the type names) After you recompile, any (logged) events that happen during a call to your function will appear grouped under a <my-func> tag in the log. For example: <my-func> <set-attribute node="element(0x8956a218)=svg:g(#g123)" name="foo" value="bar"/> </my-func> (How it works: the <my-func> tag is written from the EventTracker's constructor, and </my-func> is written from its destructor when it goes out of scope.) If you want to log fancier stuff (e.g. taking advantage of attributes) you'll currently need to write your own subclass of Debug::Event and use that instead of Debug::SimpleEvent. Note that EventTracker will not create a Debug::Event object unless the log is turned on, so you don't pay a performance penalty when you're not using it. See src/xml/simple-node.cpp for some examples (which are sadly probably more complicated than they need to be). -mental

1 0

[Fwd: SourceForge.net: Notice to project admin of upcoming project web and statistics changes]
by MenTaLguY 05 May '05

05 May '05

-----Forwarded Message----- From: SourceForge.net Team <noreply@...34...> To: mental@...3... Subject: SourceForge.net: Notice to project admin of upcoming project web and statistics changes Date: Thu, 05 May 2005 12:33:34 -0700 Greetings, This message provides an overview of upcoming changes to the SourceForge.net statistics system and the project web service. You are receiving this message because you are presently listed as an administrator of a project hosted on SourceForge.net, and we deem these changes to be important enough to notify all active project administrators. This is a one-time mailing. See the BOTTOM of this message if you are no longer active on SourceForge.net. A. Project Web Service A significant amount of recent work has gone in to improvements to the scalability and security of the project web service. On 2005-05-23, we will be remounting project group directories read-only on the project web servers (they will remain read-write from the shell server). This may impact existing project web applications. All projects are encouraged to review our new documentation and modify their application configuration as needed. http://sf.net/docs/E07/ This change is being made in accordance with the previous site mailings announcing project web changes, referencing the Master RFE at: http://sf.net/tracker/index.php? func=detail&aid=1109940&group_id=1&atid=350001 Additional details regarding the project web service, limitations, and the best ways to manage project content under this service, may be found in the recently-released Project Web, Shell and DB service guide, at: http://sf.net/docs/E07/ B. Statistics B0. Introduction and timeline The SourceForge.net team has been working for nearly a year to complete a total replacement of the existing project and site statistics infrastructure. It is our pleasure to announce that this work is reaching completion, and to provide some details to manage your expectations about the launch of the new statistics service. We expect to announce no later than 2005-05-16 the timeline for deployment of the new statistics system. This announcement will be made via the the Site Status page. https://sourceforge.net/docs/A04/ B1. Scalability, performance, outage reduction: The statistics system has been completely re-engineered as to cope with the growing size of the site user base, and of our service offering. With the launch of this new system, there are bound to be a few bugs (though we've performed extensive QA review of these changes and systems to try to reduce the problem set). We ask that all statistics related bugs found post-launch be submitted to us for review (even if they seem minor, or duplicative to those previously reported) at: https://sourceforge.net/tracker/?func=add&group_id=1&atid=200001 B2. Project ranking formula change: One of the changes which is expected to most impact projects is the alteration of the project ranking formula. This change may significantly alter the top project rankings, giving greater ranking to those projects who are not just frequently-downloaded, but also those projects who are active in using the Tracker and communication tools provided by the site. The OLD project ranking formula was: log (3 * # of forum posts for that week) + log (4 * # of tasks ftw) + log (3 * # bugs ftw) + log (10 * patches ftw) + log (5 * tracker items ftw) + log (# commits to CVS ftw) + log (5 * # file releases ftw) + log (.3 * # downloads ftw) -- This has been detailed in our older Site Documentation as seen at: https://sourceforge.net/docman/display_doc.php?docid=14040&group_id=1 With the new statistics system launch, we expect to provide new Site Documentation. The new documentation will include the final form of the NEW ranking formula. The new ranking formula splits project rankings in to three components: traffic, development, and communication. We expect the new ranking formulas to be: Traffic: ( (log(last 7 days download total) / log(highest project download total)) +(log(last 7 days logo hits) / log(highest project logo hits)) +(log(last 7 days site hits) / log(highest project site hits)) ) / 3 Development: ( (log(last 7 days cvs commit total) / log(highest project total)) +( (100-last file release age (days, max 100)) / 100 ) +( (100-last admin login age (days, max 100)) / 100 ) ) / 3 Communication: ( (log(last 7 days tracker artifact #) / log(highest project total)) +(log(last 7 days ML post count) / log(highest project total)) +(log(last 7 days forum post count) / log(highest project total)) ) / 3 Total: total = traffic + development + communication All of the 'highest' figures reference totals for the last 7 days. In the case of 'log(last 7 day site hits for this project) / log(last 7 day site hits for the project with the most site hits)', for example: If Project X is the busiest site this week with 10,000,000 hits and Project Y has had 200,000 hits then it's Y score = log(200,000) / log(10,000,000) We believe that this ranking method gives us the capability to better evaluate project activity, and to more directly reflect total product activity in how projects receive site exposure via the top site rankings. B3. Post-launch components and your feedback After the initial launch of the new statistics system, we will be working to complete the components to track CVS activity and mailing list activity. These components have not yet been completed, and are expected to launch at a later date (to be announced after the initial shake-out of the new system). Additional statistics system enhancements are also under consideration for implementation after the initial launch. Once the initial launch has occurred and you have had opportunity to examine the new facilities, we encourage you to contact us (via our Feature Request Tracker) regarding enhancements you would like to see. https://sourceforge.net/tracker/?func=add&group_id=1&atid=350001 B4. Documentation The release of new documentation (within the SourceForge.net Site Documentation collection) is expected to coincide with the new statistics system launch. Links will be provided to the new documentation from our Site Status page and Enhancements page. An itemized overview of the major changes carried by the new statistics system will be listed on the Enhancements page at time of launch. https://sourceforge.net/docs/A03/ https://sourceforge.net/docs/A04/ B5. Existing statistics data Existing statistics data (from the old statistics system) has been carried forward to the new statistics system. For the time period that the new system has actively been pulling real-time data from our servers, we plan to recompute project ranking data. All other project statistics should match those found in the old system. C. Conclusion Regarding Project Web: As documented here, project group directories will be remounted read-only on 2005-05-23. All writable files will need to be placed in the /tmp/persistent directory structure (create a subdirectory for your project). Additional details may be found in section A, above. Regarding Statistics: We are considering the completion of our new statistics system to be very high priority, and are taking only the time we need to ensure a highly consistent and functional system has been implemented. We expect to announce no later than 2005-05-16 the timeline for deployment of the new statistics system. Additional details may be found in section B, above. We hope that your team will find added benefit through our efforts, and encourage your continued support and feedback. Thank you, SourceForge.net staff http://sf.net/tracker/?func=add&group_id=1&atid=200001 --------------------------------------------------------- When the SF.net team sends out a site-wide email, we sometimes see replies that look like this: "Hey! I didn't ask to be on this list! You spammer. I hate you! I hate your dog! (insert other colorful phrases here)." The truth is, we've only sent this mailing to registered project administrators on SourceForge.net. If you do not believe you should have received this mailing, use the link above to submit a Support Request and we'll be glad to assist you. .

1 0

Memory Problems
by Richard Kowalczyk 05 May '05

05 May '05

I have a presentation of about 50 slides. When I go through the slides with inkview, it quickly eats up all of my memory (512 mb). I dont know a proper way to describe the memory usage, so I will list some info: Memory only goes up when I change slides. If it sit on the same slide, it is fine. If I just go back and forth between two slides, memory will increase each time I go back and forth, until it eventually levels off. Then I can flip back and forth between those two slides without any increase in memory. When I flip to more slides, memory goes up again. I can go through the whole presentation once, and memory will go to a certain value. Going backwards, or going to the start and showing slides again will continue to increase the memory. Eventually, after playing with the slides for about 15 minutes, inkview will crash, and print "Killed" to the console. I am basing my statements about memory on "ps aux" and "free -m". I will be glad to do anything else that anyone thinks will help explain the memory usage. I will be glad to send the presentation to anyone also. It is about 7 mb uncompressed. Thanks, Rich. -- Richard Kowalczyk <richk@...805...>

1 0

layer add hang
by MenTaLguY 05 May '05

05 May '05

The layer addition hang appears to be due to the failure of the "change handlers should only be called for events that happen after they are registered" invariant I mentioned. Basically what is happening is that on a layer switch, the layer selectors each register a new handler (after unregistering the old one), but those handlers immediately get called for the event that has already happened, repeating the process and creating an infinite treadmill. The obvious fix is to defer addition of new change handlers to the list until after all the existing handlers have been called. Of course that is not a sufficient guarantee for the recursive case, but hopefully there are no recursive changes involved here. -mental

1 0