- Dillo-dev - lists.mailman3.com

Re: [Dillo-dev] FLTK

by Alex Pavloff

> What *does* take far more time is the initial rendering, > while rewrapping is exactly as fast as before. This (and also > that Pango is only used for drawing, and determing the size > of fonts and words) lets me assume that Dw_page_shape_glyphs > is not very efficient. I've tried to track down this with > gprof, but somehow the profile does not give me much clues. > Perhaps it is worth to optimize this code. (The comparison is > a bit unfair because of the different grades of matureness :-) Hi guys, It seems obvious to me, that before an effort is made to rip out one widget system and replace it with another widget system, a little more profiling is in order to confirm indeed, that the problem is Pango. If Pango is the problem, again, before changing the underlying widget system, perhaps some investigation in concert with the Pango developers should be done to see if there's some way to optimize dillo or Pango. Just my $0.02. Alex Pavloff - apavloff(a)esatechnology.com ESA Technology ---- www.esatechnology.com ------- Linux-based industrial HMI ------ -------- www.esatechnology.com/5k -------

Oct. 16, 2003

Re: [Dillo-dev] [PATCH] bug in menu.c (SOLVED)

by Jorge Arellano Cid

On Thu, 16 Oct 2003, Frank de Lange wrote: > [...] > Found the cause. This error is present in the CVS version of Dillo as > well, but there it does not matter much (probably why it was not > discovered before). In menu.c:Menu_add(): > > if (callback != NULL) { > gtk_signal_connect(GTK_OBJECT(menuitem), "activate", > (GtkSignalFunc) Menu_pick_url, data); > ^^^^ > > "data" should be "bw", as Menu_pick_url() needs a BrowserWindow* as > second parameter. In CVS Dillo "data" (which is used as user data for > callbacks) happens to be always set to "bw", so this bug did not > manifest itself. In the patched version, some callbacks expect a > DilloDoc* as parameter though... So Menu_pick_url() is called with a > DilloDoc* instead of a BrowserWindow*... > > The fix is simple: > > <PATCH> > diff -pruN dillo/src/menu.c dillo_patched/src/menu.c > --- dillo/src/menu.c 2003-09-25 15:10:24.000000000 +0200 > +++ dillo_patched/src/menu.c 2003-10-16 14:56:38.733763968 +0200 > @@ -109,7 +109,7 @@ static GtkWidget * > } > if (callback != NULL) { > gtk_signal_connect(GTK_OBJECT(menuitem), "activate", > - (GtkSignalFunc) Menu_pick_url, data); > + (GtkSignalFunc) Menu_pick_url, bw); > gtk_signal_connect(GTK_OBJECT(menuitem), "activate", > (GtkSignalFunc) callback, data); > } > </PATCH> > > Jorge, can you commit this patchlet to CVS? Done! Cheers Jorge.-

Oct. 16, 2003

Re: [Dillo-dev] [PATCH] version 1 of keyboard navigation patch

by Glyn Kennington

Frank de Lange wrote: > > [segfault involving links in alt tags] > I have not been able to reproduce this yet, will try again and fix > anything I find... But, as stated, it does not happen here... Could have > something to do with your next remark? > > [patches not applying cleanly] Well, as no-one else seemed affected, I tried again with a couple more timestamps on cvs. I found a tree to which the patch would apply cleanly (no fuzz factor, no offsets), and yes, the bug had disappeared. (My broken version also didn't shade the images when they were selected - a feature I didn't know I was missing.) I'll take a closer look at some of the cvs changes before my broken version, but it's probably nothing serious. Glyn -- glyn.kennington(a)ox.compsoc.net

Oct. 16, 2003

[PATCH] bug in menu.c (SOLVED)

by Frank de Lange

Stephan Goetter wrote: >Hi, > >I just got this, after opening a new tab, going to >http://www.foobar123.org/ (yes, it doesn't exist :) >and then trying to close the Tab. > >Program received signal SIGSEGV, Segmentation fault. >Menu_pick_url (widget=0x8158e80, client_data=0x8113f20) at menu.c:112 >112 gtk_object_remove_data(GTK_OBJECT (bw->menu_popup.over_image), "url2"); >(gdb) backtrace full >#0 Menu_pick_url (widget=0x8158e80, client_data=0x8113f20) at menu.c:112 > bw = (struct _BrowserWindow *) 0x8113f20 >#1 0x40142ef6 in gtk_marshal_NONE__NONE () from /usr/lib/libgtk-1.2.so.0 >No symbol table info available. >#2 0x40187e16 in gtk_signal_remove_emission_hook () from /usr/lib/libgtk-1.2.so.0 >No symbol table info available. >#3 0x40186f3b in gtk_signal_handlers_destroy () from /usr/lib/libgtk-1.2.so.0 >No symbol table info available. >#4 0x40184b8e in gtk_signal_emit () from /usr/lib/libgtk-1.2.so.0 >No symbol table info available. >... > >Greets, >Stephan > Found the cause. This error is present in the CVS version of Dillo as well, but there it does not matter much (probably why it was not discovered before). In menu.c:Menu_add(): if (callback != NULL) { gtk_signal_connect(GTK_OBJECT(menuitem), "activate", (GtkSignalFunc) Menu_pick_url, data); ^^^^ "data" should be "bw", as Menu_pick_url() needs a BrowserWindow* as second parameter. In CVS Dillo "data" (which is used as user data for callbacks) happens to be always set to "bw", so this bug did not manifest itself. In the patched version, some callbacks expect a DilloDoc* as parameter though... So Menu_pick_url() is called with a DilloDoc* instead of a BrowserWindow*... The fix is simple: <PATCH> diff -pruN dillo/src/menu.c dillo_patched/src/menu.c --- dillo/src/menu.c 2003-09-25 15:10:24.000000000 +0200 +++ dillo_patched/src/menu.c 2003-10-16 14:56:38.733763968 +0200 @@ -109,7 +109,7 @@ static GtkWidget * } if (callback != NULL) { gtk_signal_connect(GTK_OBJECT(menuitem), "activate", - (GtkSignalFunc) Menu_pick_url, data); + (GtkSignalFunc) Menu_pick_url, bw); gtk_signal_connect(GTK_OBJECT(menuitem), "activate", (GtkSignalFunc) callback, data); } </PATCH> Jorge, can you commit this patchlet to CVS? Cheers//Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Oct. 16, 2003

Re: [Dillo-dev] [PATCH] version 1 of keyboard navigation patch

by Victor Sahlstedt

On Thu, 16 Oct 2003 13:13:04 +0200 Frank de Lange <frank(a)unternet.org> wrote: > I have not been able to reproduce this yet, will try again and fix > anything I find... But, as stated, it does not happen here... Could have > something to do with your next remark? > > [ LIST: has anyone else seen this? ] No, I can't reproduce this. - Victor

Oct. 16, 2003

Re: [Dillo-dev] [PATCH] version 1 of keyboard navigation patch

by Frank de Lange

Glyn Kennington wrote: > I like most of it. But after a very short time, I've found some things > >that I think need improving. So, in order of decreasing importance: > >1) I've found a bug that will cause dillo to crash: > > Go to a page whose first link is in an alt-tagged image. (The one I've > been using for this bugreport is www.guardian.co.uk .) > > Press the key for `link DOWN'. This will make the image link the > current link (status bar reads "http://www.guardian.co.uk/0,6961,,00.html") > > Go to another page. This can be by clicking Home, Back etc, typing in > another URL, or activating a link on the current page. > > Once dillo is about to start rendering the new page, it dies with a > segfault. gdb tells me that it's at the end of Dw_image_destroy(): > g_free (image->alt_text); > >This crash still occurs if I move to the next link (outside the image) >before going to a different page. > > I have not been able to reproduce this yet, will try again and fix anything I find... But, as stated, it does not happen here... Could have something to do with your next remark? [ LIST: has anyone else seen this? ] >2) Patching problems. I know the email said `use cvs 14/10/2003', but >the patch filename includes the string 20031015, so I attempted to patch >against my just-updated checkout. This normally wouldn't be a problem >over the course of only a day, but today there was some change to the >arguments of a_Commands_set_selection which broke the patch. If you >could include on your webpage the exact date and time of the CVS >checkout, near the link to the patch, it would probably prevent this >sort of confusion from happening again. > > Hmmm... tricky. I'll add a GMT date/time to the CVS version to be used. The date-string on the patch is autogenerated by my patch scripts, it just tells the date I made a release. It is not necessarily related to the CVS date the patch is based on. Sometimes I make incremental changes to a patch without updating CVS. >3) The TODO mentions more bindings. Do you intend to make Back > bindable? > > I intend to make everything bindable. That is, everything which can be bound. That is 'the GTK+ way' of doing things... Cheers//Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Oct. 16, 2003

Re: [Dillo-dev] [PATCH] version 11 of tab/frame patch, added keyboard navigation

by Frank de Lange

Stephan Goetter wrote: >Hi, > >I just got this, after opening a new tab, going to >http://www.foobar123.org/ (yes, it doesn't exist :) >and then trying to close the Tab. > >Program received signal SIGSEGV, Segmentation fault. >Menu_pick_url (widget=0x8158e80, client_data=0x8113f20) at menu.c:112 >112 gtk_object_remove_data(GTK_OBJECT (bw->menu_popup.over_image), "url2"); >(gdb) backtrace full >#0 Menu_pick_url (widget=0x8158e80, client_data=0x8113f20) at menu.c:112 > bw = (struct _BrowserWindow *) 0x8113f20 >#1 0x40142ef6 in gtk_marshal_NONE__NONE () from /usr/lib/libgtk-1.2.so.0 >No symbol table info available. >#2 0x40187e16 in gtk_signal_remove_emission_hook () from /usr/lib/libgtk-1.2.so.0 >No symbol table info available. >#3 0x40186f3b in gtk_signal_handlers_destroy () from /usr/lib/libgtk-1.2.so.0 >No symbol table info available. >#4 0x40184b8e in gtk_signal_emit () from /usr/lib/libgtk-1.2.so.0 >No symbol table info available. >... > >Greets, >Stephan > > Hmmm, this did not happen here, but that was caused by me using a proxy (privoxy) which produces a nice 'not found' page... Will look in to it asap... Cheers//Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Oct. 16, 2003

Re: [Dillo-dev] [PATCH] version 11 of tab/frame patch, added keyboard navigation

by Frank de Lange

johannes leimbach wrote: >Hi! > >You're patch works fine, thanks for the good work. >One (minor) bug though: > >It seems to mess up the tab order of the elements. >Take for example an online forum (phpbb). Go to "Login", so you must >enter an username+password. Before the patch (in 0.7.3) i could set the >focus from textbox 2 textbox with TAB, now it "jump" to a (for me) random >element. > > Yeah, Tab is a bit weird right now. I'll probably bind Tab to gtk_contain_focus (the way it was) in the next version, both for the tab/frame and non-frame version of the patch.. Cheers//Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Oct. 16, 2003

Re: [Dillo-dev] Problem: Dillo not rendering /. correctly

by Ralph Slooten

On Wed, 15 Oct 2003 16:36:17 +0200 johannes leimbach <johannes.leimbach(a)gmx.de> wrote: > Do you have any ideas what might cause this visual annoyance? Yeah, many errors on /.org ... too many. I files a bug report there which just went unnoticed and was marked within 1 day as "closed", even though there was no difference. Due to this problem, and the fact that I love Dillo and slashdot, I created my own site which downloads (using a local cookie) and renders slashdot (front page only) through tidy. If the main page is what you are looking for, you can use my site (works ONLY for Dillo, rejects other browsers totally) at http://slashdot.axljab.homelinux.org/ Greetings Ralph -- http://axljab.homelinux.org/ "...the software said Win95 or better, so I installed Linux"

Oct. 16, 2003

Re: [Dillo-dev] FLTK

by Melvin Hadasht

Hi, > How do I turn off anti-aliasing? I'd like to try it again if you think this > may speed-up redraws. add something like this in your ~/.fonts.conf <match target="font"> <test qual="any" name="size" compare="more"> <int>7</int> </test> <test qual="any" name="size" compare="less"> <int>10</int> </test> <edit name="antialias" mode="assign"> <bool>false</bool> </edit> </match> This will turnoff AA for sizes between 7 and 10 points Cheers -- Melvin Hadasht

Oct. 16, 2003