[Dillo-dev]Re: Dillo configuration (fwd) From: Tomas - 2002-03-31 12:45 Attachments: Message as HTML On Wed, Mar 27, 2002 at 10:23:20AM -0400, Jorge Arellano Cid wrote: > ---------- Forwarded message ---------- > Date: Tue, 5 Feb 2002 16:04:30 -0300 (CLST) > From: Jorge Arellano Cid > To: Chris Wareham > Subject: Re: Dillo configuration =2E.. >=20 > * There's a third history mechanism that hasn't made it into > the code yet: linear history. It's designed to provide a > chronological list of visited urls, with its own button a the > leftmost part. The problem with it is that it should provide > scrolling, highlighting of the current-page-item inside the > history list and display the url at the status bar. > A nice solution would be to have a sort of popup that can be > constrained to a certain number of items, with scrolling, > highlighting and url-showing (so it can be used for > back/Fwd/linear history widgets). > If you can do that, it'd be very helpful. Just tell me and I > send you the code I have. Could you send me the code you have already done, did you ask in gtk-list if there is a widget like this or should i do that. greets Tomas [Dillo-dev][PATCH] block non local browsing From: Melvin Hadasht - 2002-03-30 21:13 Hi, Finally, I first made a patch to block all non local (non 127.x.x.x) requests. So browsing files or http pages on the localhost is still permitted. Non local (dns) requests are reported unknown (dns returns 0). This behavior is available using -l (for local) command line option. This feature should be used when browsing untrusted pages. If there is an interest, I'll try to make it a run-time option. Patch (4k) against cvs (cvs diff -c) is located at: http://melvin.hadasht.free.fr/dillo/dillo-local-cvs-30-03-2002.patch and http://freefluid.dyndns.org/dillo/dillo-local-cvs-30-03-2002.patch Next step is providing remote control to dillo. Cheers. -- Melvin Hadasht [Dillo-dev]New site link From: Calvin Arndt - 2002-03-30 21:09 dillo.s...net site needs to be updated with a link to the new dillo.cipsga.org.br site. New site is very hard to find (Even google search does not return a link!) Re: [Dillo-dev]Bookmarks menuitem in state bar and scroll menu From: Jorge Arellano Cid - 2002-03-30 14:06 Tomas, > One of my doubts before doing it was if i should patch against > 0.6.4 or against the latest CVS code. I never worked before with CVS and > diff so if i did something wrong just tell me so the next time i will > try to do it better :) > > What i did was: > > - get the code from CVS in ~/dillo/dillo-cvs > - copy to another directory ~/dillo/dillo-tomas > - work on ~/dillo/dillo-tomas > - diff -pru dillo dillo-tomas (i saw the -pru flags for diff in another > patch) That's exactly the way I do it! BTW, I also put the '-pru' flags in the patching section of the new site (it was somewhat hidden in a README file; funny but true! ;) Cheers Jorge.- Re: [Dillo-dev]Bookmarks menuitem in state bar and scroll menu From: Tomas - 2002-03-30 00:49 Attachments: Message as HTML On Fri, Mar 29, 2002 at 10:07:15AM -0400, Jorge Arellano Cid wrote: >=20 >=20 > On Thu, 28 Mar 2002, Tomas wrote: >=20 > > I have added scroll to the bookmark menu and the status bar now reflects > > the url is pointed with the mouse as in Back/Forwd history. > > > > The problem i see is when the menus are big enough to do scroll > > then covers the state bar so it's not posible to see the address. > > > > This is my first patch so any sugestion is very apreciated :) it's > > with the CVS code. ^^^^ Sorry i was trying to say that the patch i did was against the CVS code from that day and not against the stable version (0.6.4) One of my doubts before doing it was if i should patch against 0.6.4 or against the latest CVS code. I never worked before with CVS and diff so if i did something wrong just tell me so the next time i will=20 try to do it better :)=20 What i did was: - get the code from CVS in ~/dillo/dillo-cvs - copy to another directory ~/dillo/dillo-tomas - work on ~/dillo/dillo-tomas - diff -pru dillo dillo-tomas (i saw the -pru flags for diff in another patch) =20 >=20 > The code is not in CVS now, as we're still working on it with > Tomas. The idea is to make a generic popup that allows scrolling, > shows the URL in the status bar, and that has a customizable > number of displayed-entries. >=20 >=20 > Cheers > Jorge.- >=20 [Dillo-dev]Re: CSS From: jonathan chetwynd - 2002-03-28 22:22 connected fine, comments please on CSS1 and CSS2 which do not seem to be supported at present... will they be? and if so when-ish? really its a fabulous browser - good stuff thanks again [Dillo-dev]CSS From: jonathan chetwynd - 2002-03-28 21:58 Does Dillo support CSS1 or CSS2? if not is there a planned for day. I've no idea of the workload, or bloat factor, but would like to use it. I'm busy rebuilding our website, and can't just get dillo to connect to the internet. damn, ipaq and ssh is fine, as is the web connection, they just wont talk to each other. oh well maybe tomorrow. thanks Re: [Dillo-dev]Bookmarks menuitem in state bar and scroll menu From: Lars Clausen - 2002-03-28 18:05 On Thu, 28 Mar 2002, tomas@pu... wrote: > Hi all, >=20 > I have added scroll to the bookmark menu and the status bar now reflects > the url is pointed with the mouse as in Back/Forwd history. Cool, my bookmarks are getting to that state soon. If you feel like hacking more at the bookmarks, I'd love to see submenus in the bookmarks. Galeon does a good job of handling them, if you want inspiration.=20 -Lars --=20 Lars Clausen (http://shasta.cs.uiuc.edu/~lrclause)| H=E5rdgrim of Numenor "I do not agree with a word that you say, but I |------------------------= ---- will defend to the death your right to say it." | Where are we going, and --Evelyn Beatrice Hall paraphrasing Voltaire | what's with the handbas= ket? [Dillo-dev]Bookmarks menuitem in state bar and scroll menu From: Tomas - 2002-03-28 15:48 Attachments: Message as HTML Message as HTML [Dillo-dev]patch: embedding dillo From: Melvin Hadasht - 2002-03-28 00:03 Hi, Thanks for this great piece of software. I use it all day (as long as the pages do not require java/javascript). It is really amazing fast and renders correctly the majority of the sites I visit. Please continue such a good work! I came to the idea that dillo can be used as an embedded html viewer in email MUA, for example. So I used the Gtk_Socket/Gtk_Plug feature to make a very small patch that allows dillo to be embedded in an existing gtk_socket window. It adds a '-xid' command line option (should be placed after the URI) which need the xid of a window (provided by the GDK_WINDOW_XWINDOW(GdkWindow *) macro). That way, any GTK application can create a gtk_socket, then it launches dillo with the -xid option and the first dillo window will be embedded. This worked so far quiet good: I hacked my Sylpheed-Claws version to support it. The dillo patch against cvs is at: http://melvin.hadasht.free.fr/dillo/dillo-embed-cvs-27-03-2002.patch or at http://freefluid.dyndns.org/dillo/dillo-embed-cvs-27-03-2002.patch (the latter being not up 24/24). But as html mail is so dangerous, I wanted to ask you: is there any plan to make dillo able to run in offline mode? For example by using a command line. To continue with my plan to use dillo as a versatile tool, I plan to make another patch where dillo will be able to listen on a Unix socket for commands sent by other processes. These commands would be very simple: - open an URL in current window, - open an URL in a new window (with the xid supplied, xid = 0 being in a new GTK_TOP_LEVEL) - close window xid (!= 0), or quit (xid = 0). The patch should not be big (at first sight), but I don't know when I will have time to do it. So, is there any interest in such features? Any suggestions or comments to the patch or to my ideas are welcome! Thanks. -- Melvin Hadasht Re: [Dillo-dev][RFC] Pixmap (XPM) image support / Broken pixmap From: Sebastian Geerken - 2002-03-26 18:21 Hi, On Mon, Mar 25, 2002 at 03:50:11PM -0800, Michael Taht wrote: > I don't know if this has been discussed before, but why not bite the > bullet and evolve up to the gdk-pixbuf library? I'd second Livio, simply for the reason that the the current code works (rather) well, and that it is simple enough to maintain, but otherwise gives more control over it. On Mon, Mar 25, 2002 at 09:39:00PM -0300, Livio Baldini Soares wrote: > Howdy Dillo people, > > This patch is an incremental patch to my previous patch which adds > pixmap (XPM) support to Dillo (see: > http://www.geocrawler.com/lists/3/SourceForge/702/0/8209368/). > > It relies on that support to effectively display a "404" pixmap on > 404 or redirections for non-root URLs. It basically adds a pixmap in > src/xpm.h called notfound_xpm. I tried to do frowning face by hand (I > know it's ugly, but it's just for testing). You can replace it with > whichever XPM you prefer. > > But the "real" changes are located in the cache.c, they're only > about 5 lines "setting up" the entry data. The patch is at: > > http://www.ime.usp.br/~livio/dillo/patches/notfound-pixmap.diff > > As always, please send feedback! I planned to reuse the code which reads the image data (anything before DwImage) for background images, and I doubt that such a background is pleasant for the user ;-) Rather, backgrounds should then simply not be seen (perhaps, the user could silently be notified). An idea I had already, was to implement the display of failures in DwImage. Some time ago, I wrote something about an interface, which replaces (mostly) DwImage in the image code which does not actually display images, and is implemented by DwImage and the code handling backgrounds. An idea is to add a failed message, which is called for any broken images (40x, corrupt data), which is ignored by backgrounds, but puts an DwImage into a state, in which is showed that an error has occurred. Sebastian Re: [Dillo-dev][RFC] Pixmap (XPM) image support [Patch 1/3] From: Pekka Lampila - 2002-03-26 17:17 On Mon, 25 Mar 2002 22:09:26 -0300 Livio Baldini Soares wrote: > Does anyone know here "de facto" if gdk-pixbuf does progressive > loading of images from an in-memory buffer? Or of another "well > established" light library which does this? http://developer.gnome.org/doc/API/gdk-pixbuf/gdkpixbufloader.html -- Pekka Lampila * medar@ka... * If pro is the opposite of con, http://kirjasto.org/medar * what is the opposite of progress? Re: [Dillo-dev][RFC] Pixmap (XPM) image support [Patch 1/3] From: Livio Baldini Soares - 2002-03-26 01:09 Hello Michael! Michael Taht writes: > I don't know if this has been discussed before, but why not bite the > bullet and evolve up to the gdk-pixbuf library? Well to tell you the truth, I have been giving thought about this issue too... to use an external image library, like gdk-pixbuf or imlib, etc. I see advantages and disadvantages to this, but I haven't really convinced myself which is better. Some people seem to like that Dillo does not depend on big graphics libraries. On the other hand, maintaining various image codes is not very good for development... What are opinions from other developers? Jorge? Sebastian? > I believe it supports progressive loading of a variety of image > types. Well _this_ (IMHO) is the main reason for not changing to an image library. I haven't seen any library which does progressive in-memory loading of any type of images. This information can be wrong, but look at the gdk-pixbuf/io-xpm.c comment (from GTK 2.0.0): /* * FIXME xpm loading progressively is not properly implemented. * Instead we will buffer to a file then load that file when done. * This is very broken but it should be relayively simple to fix * in the future. */ (Which is quite horrible.. they use a temporary file in /tmp/... :( Anyway, my knowledge of most image libraries is _very_ limited, so I'll stop guessing and ask: Does anyone know here "de facto" if gdk-pixbuf does progressive loading of images from an in-memory buffer? Or of another "well established" light library which does this? If we find a good candidate, it might help out our discussion. best regards, -- Livio [Dillo-dev][RFC] Broken pixmap on 404 non-root URLs [Patch 2/3] From: Livio Baldini Soares - 2002-03-26 00:39 Howdy Dillo people, This patch is an incremental patch to my previous patch which adds pixmap (XPM) support to Dillo (see: http://www.geocrawler.com/lists/3/SourceForge/702/0/8209368/). It relies on that support to effectively display a "404" pixmap on 404 or redirections for non-root URLs. It basically adds a pixmap in src/xpm.h called notfound_xpm. I tried to do frowning face by hand (I know it's ugly, but it's just for testing). You can replace it with whichever XPM you prefer. But the "real" changes are located in the cache.c, they're only about 5 lines "setting up" the entry data. The patch is at: http://www.ime.usp.br/~livio/dillo/patches/notfound-pixmap.diff As always, please send feedback! The next patch should be the unhandled MIME type patch... but it might be ready only for tomorrow. kind regards! -- Livio Re: [Dillo-dev][RFC] Pixmap (XPM) image support [Patch 1/3] From: Michael Taht - 2002-03-25 23:50 I don't know if this has been discussed before, but why not bite the bullet and evolve up to the gdk-pixbuf library? I believe it supports progressive loading of a variety of image types. On Mon, 2002-03-25 at 15:02, Livio Baldini Soares wrote: > Hello Dillo people! > > On my quest to squash bug #81 (see > http://www.geocrawler.com/lists/3/SourceForge/702/0/8169781/ for > previous post), Andreas suggested I try to combine my fix with bug > #245 (broken images). So I have adopted that bug too, and will try to > kill them all with a bottom-up strategy. > > First of all, I needed a generic way to insert an image inside > Dillo's page from within the cache module ('cause the cache is the one > which will detect missing images/unhandled mime types, etc). I thought > of various ways to do this, but the most natural was using a pixmap > (XPM) renderer. This is my motivation: _internal use_. > > I have passed the last weekend implementing this. Basically I > started from GTK's 1.2.10 gdkpixmap.c and GTK's 2.0.0 > gdk-pixbuf/io-xpm.c. There were two problems there. First there was a > lot of stuff which could be simplified for Dillo's use. I've tried to > do this and I think only two functions are still similar to the GTK's > version. > > The second (and most important and troublesome), was that GTK's > pixmap render is synchronous, ie, it must have _all_ the data before > it can start rendering. I've made it possible to do a progressive > render. > > So, now dillo has support for . I know that noone > actually uses xpm's out there, and probably no other browser > implements it, but I figured since I implemented for internal use, why > not do it "right" and make it robust/progressive. > > The patch is at (against CVS from 25-Mar-2002): > http://www.ime.usp.br/~livio/dillo/patches/pixmap-support.diff > > There is a test page here: > http://www.ime.usp.br/~livio/dillo/xpm-test/ > > The first two images will not appear (there are purposely broken to > test the next patch I'll be sending to the list), but the next 5 xpms > should render perfectly. > > Does anybody see a better way of doing this in Dillo than what I > proposed? Comments are always very welcome! > > PS: There seems to be a _very_ rare "glitch" in my code. I can't > reproduce it, but sometimes the XPM is rendered incorrectly... by > pressing Back and Forward, or Reload, it re-renders fine. I stared at > the code for about 2 hours and couldn't see what's causing it. If > anyone sees this problem and has either a fix or a reproducible case, > please contact me! > > best regards to all, > > -- > Livio > > _______________________________________________ > Dillo-dev mailing list > Dillo-dev@li... > https://lists.so....net/lists/listinfo/dillo-dev [Dillo-dev][RFC] Pixmap (XPM) image support [Patch 1/3] From: Livio Baldini Soares - 2002-03-25 23:02 Hello Dillo people! On my quest to squash bug #81 (see http://www.geocrawler.com/lists/3/SourceForge/702/0/8169781/ for previous post), Andreas suggested I try to combine my fix with bug #245 (broken images). So I have adopted that bug too, and will try to kill them all with a bottom-up strategy. First of all, I needed a generic way to insert an image inside Dillo's page from within the cache module ('cause the cache is the one which will detect missing images/unhandled mime types, etc). I thought of various ways to do this, but the most natural was using a pixmap (XPM) renderer. This is my motivation: _internal use_. I have passed the last weekend implementing this. Basically I started from GTK's 1.2.10 gdkpixmap.c and GTK's 2.0.0 gdk-pixbuf/io-xpm.c. There were two problems there. First there was a lot of stuff which could be simplified for Dillo's use. I've tried to do this and I think only two functions are still similar to the GTK's version. The second (and most important and troublesome), was that GTK's pixmap render is synchronous, ie, it must have _all_ the data before it can start rendering. I've made it possible to do a progressive render. So, now dillo has support for . I know that noone actually uses xpm's out there, and probably no other browser implements it, but I figured since I implemented for internal use, why not do it "right" and make it robust/progressive. The patch is at (against CVS from 25-Mar-2002): http://www.ime.usp.br/~livio/dillo/patches/pixmap-support.diff There is a test page here: http://www.ime.usp.br/~livio/dillo/xpm-test/ The first two images will not appear (there are purposely broken to test the next patch I'll be sending to the list), but the next 5 xpms should render perfectly. Does anybody see a better way of doing this in Dillo than what I proposed? Comments are always very welcome! PS: There seems to be a _very_ rare "glitch" in my code. I can't reproduce it, but sometimes the XPM is rendered incorrectly... by pressing Back and Forward, or Reload, it re-renders fine. I stared at the code for about 2 hours and couldn't see what's causing it. If anyone sees this problem and has either a fix or a reproducible case, please contact me! best regards to all, -- Livio Re: [Dillo-dev]OpenSSL/Cookies Patch for Dillo 0.6.4 From: Willem van Engen - 2002-03-25 14:19 Thanks for the patch! However, my configure fails; there seems to be an error in the automake/autoconf generated configure. I tried to hack around this one error, but there are many other errors. So the generated configure looks really broken. What version of autoconf and automake did you use? I currently have automake 1.5 and autoconf 2.53. - Willem van Engen (After successful patching) /usr/ports/www/dillo/work/dillo-0.6.4# uname -a FreeBSD foo 4.5-STABLE FreeBSD 4.5-STABLE #1: Sat Feb 2 06:36:47 CET 2002 wvengen@foo:/usr/obj/usr/src/sys/FOO i386 /usr/ports/www/dillo/work/dillo-0.6.4# automake automake: configure.in: required file `./depcomp' not found /usr/local/share/automake/am/depend2.am: AMDEP does not appear in AM_CONDITIONAL /usr/local/share/automake/am/depend2.am: AMDEP does not appear in AM_CONDITIONAL /usr/local/share/automake/am/depend2.am: AMDEP does not appear in AM_CONDITIONAL /usr/local/share/automake/am/lang-compile.am: AMDEP does not appear in AM_CONDITIONAL /usr/local/share/automake/am/depend2.am: AMDEP does not appear in AM_CONDITIONAL /usr/local/share/automake/am/depend2.am: AMDEP does not appear in AM_CONDITIONAL /usr/local/share/automake/am/depend2.am: AMDEP does not appear in AM_CONDITIONAL /usr/local/share/automake/am/lang-compile.am: AMDEP does not appear in AM_CONDITIONAL /usr/ports/www/dillo/work/dillo-0.6.4# autoconf /usr/ports/www/dillo/work/dillo-0.6.4# ./configure --includedir="/usr/local/include" --libdir="/usr/local/lib" --with-jpeg-inc="/usr/local/include" --enable-ssl checking for a BSD-compatible install... /usr/bin/install -c checking whether build environment is sane... yes checking whether make sets ${MAKE}... yes checking for working aclocal... found checking for working autoconf... found checking for working automake... found checking for working autoheader... found checking for working makeinfo... found checking for gcc... gcc checking for C compiler default output... a.out checking whether the C compiler works... yes checking whether we are cross compiling... no checking for suffix of executables... checking for suffix of object files... o checking whether we are using the GNU C compiler... yes checking whether gcc accepts -g... yes checking for gcc option to accept ANSI C... none needed checking for ranlib... ranlib checking for gtk-config... /usr/X11R6/bin/gtk-config checking for GTK - version >= 1.2.0... yes checking for jpeg_destroy_decompress in -ljpeg... yes checking how to run the C preprocessor... gcc -E checking for ANSI C header files... yes checking for sys/types.h... yes checking for sys/stat.h... yes checking for stdlib.h... yes checking for string.h... yes checking for memory.h... yes checking for strings.h... yes checking for inttypes.h... yes checking for stdint.h... no checking for unistd.h... yes checking jpeglib.h usability... yes checking jpeglib.h presence... yes checking for jpeglib.h... yes checking jconfig.h usability... yes checking jconfig.h presence... yes checking for jconfig.h... yes checking jerror.h usability... yes checking jerror.h presence... yes checking for jerror.h... yes checking jmorecfg.h usability... yes checking jmorecfg.h presence... yes checking for jmorecfg.h... yes ./configure: line 3439: syntax error near unexpected token `;}' ./configure: line 3439: `LIBS="-lssl rejected by the preprocessor!" >&2;}' /usr/ports/www/dillo/work/dillo-0.6.4# cat configure | head -n 3445 | tail -n 10 echo $ECHO_N "checking for SSL_connect in -lssl... $ECHO_C" >&6 if test "${ac_cv_lib_ssl_SSL_connect+set}" = set; then echo $ECHO_N "(cached) $ECHO_C" >&6 else ac_check_lib_save_LIBS=$LIBS LIBS="-lssl rejected by the preprocessor!" >&2;} { echo "$as_me:$LINENO: WARNING: $ac_header: proceeding with the preprocessor's result" >&5 echo "$as_me: WARNING: $ac_header: proceeding with the preprocessor's result" >&2;};; no:yes $LIBS" cat >conftest.$ac_ext <<_ACEOF /usr/ports/www/dillo/work/dillo-0.6.4# Re: Re: [Dillo-dev]download behavior From: Paul Chamberlain - 2002-03-24 23:12 From Mark Schreiber ----- > * Use MIME types, file extensions, *and* protocol names to specify the > helper program. plural... I think I should be able to specify several handlers for a mime type, file extension, etc. and have the menu offer each of them. Sometimes I want to view it, sometimes I want to edit it, sometimes I want to print it, etc. In particular, I have several image-viewers with different strengths. -- Paul Chamberlain, tif@ti... [Dillo-dev]OpenSSL/Cookies Patch for Dillo 0.6.4 From: Jonathan P Springer - 2002-03-24 15:20 At the request of a Dillo user, I've posted my OpenSSL/Cookies patch for version 0.6.4 at http://members.bellatlantic.net/~vze2mqqv/dillo-ssl-cookies.0.6.4.diff This is not a patch for the faint of heart. To deal with SSL Proxy Tunnelling, I essentially had to rewrite the I/O engine. I will not be maintaining this patch against CVS versions, as I believe Jorge has his own plans for SSL. If that's not the case, I'll recreate it for the next major release. To apply the patch, download it into your dillo source root directory and... patch -p1 <./dillo-ssl-cookies.0.6.4.diff automake autoconf ./configure --enable-ssl make ./src/dillo https://www.so....net/ # simple test case make install Best of luck. -js -- -Jonathan P Springer ------------------------------------------------------------------------------ "A standard is an arbitrary solution to a recurring problem." - Joe Hazen Re: [Dillo-dev]download behavior From: Stephen Lewis - 2002-03-23 20:56 On Sat, 23 Mar 2002 15:40:53 -0500 Mark Schreiber wrote: > * Be able to download a file and then hand it to a helper program *or* > just pass the URL as an argument to the helper program. I think rather than having an exclusive choice like that, it'd be nice to have two separate modules - 'download', and 'post-process'. This would let you use, say, wget to download, and then still be able to execute an external app after its done downloading based on its mime-type to display the downloaded file. Stephen Lewis Re: [Dillo-dev]download behavior From: Mark Schreiber - 2002-03-23 20:41 On Fri, Mar 22, 2002 at 08:06:54AM -0500, Imad Hussain wrote: > On Fri, 22 Mar 2002 13:24:45 +0100 > Tomas wrote: > > > I just wondering if this behaviour is what you want or is going to be > > change in the future. > > > > Also ask if there is any planning of implement resume for file > > downloading. > > Y'know, it would be pretty nice if you could force dillo to use an > external utility (e.g. wget) to take care of file downloads completely. > I'd imagine the changes necessary to the source would not be monumental > in scope -- and for those without wget (or whatever), this could be > controlled by an option in .dillorc. It takes much of the burden off of > dillo itself; after all, why reinvent the wheel? A navigator style interface would be really nice, with "helpers". I've wished for the ability to make mailto: links open in mutt for some time. I'd suggest that such a system: * Be able to download a file and then hand it to a helper program *or* just pass the URL as an argument to the helper program. * Be able to hand components of URLs to external programs. If you take a look in Navigator's preferences dialog, you'll see lines like telnet rxvt -e telnet %h %p This lets navigator hand just the host and the port off. Since a lot of programs can't take a whole URL, this is really useful. %h host %u username %p port %s download file and replace with name of local file * Use MIME types, file extensions, *and* protocol names to specify the helper program. * Use /etc/mime.types if present to obtain mime type/extension mappings. * To improve startup speed (Navigator starts slowly because it has to parse zillions of files each startup), it might be worthwhile to do all the parsing work lazily -- don't do it for an instance of dillo until dillo runs across something it doesn't understand how to handle locally. -- Best of luck, Mark Schreiber Re: [Dillo-dev]navigation feature From: Mark Schreiber - 2002-03-23 19:25 On Thu, Mar 21, 2002 at 06:05:51PM -0600, John Utz wrote: > i sent a patch in months ago to do the 'back' part. > > but it never showed up in the release (dillo-6.3). > > sort of takes the wind out of one's sails when it comes to contributing > patches..... Speaking of which, is there any reason that the dillo project avoids using the sourceforge patch/bug tracking engines? I'll grant that the bug tracking engine it does use is fast and lightweight...but the sourceforge system does have some nice features: * I really wish I could follow-up on bugs with questions or comments. The SF bug tracker lets people view bugs and comment on them. Jorge has stated that he'd like to keep this ML low traffic -- which means not sending all replies about patches to the list. However, if you reply directly to the person, you do not allow anyone else to identify problems with your comments or answer your questions. SF's bug tracker could handle this. * It feels more rewarding to have a patch sitting in the patch manager than to email it away and not see it again until it gets in the source. * Dropping patches in a patch manager lets other people possibly identify problems without Jorge having to run through them all first. * I accidently sent a partial patch to Jorge some time ago. Since I was using email, I had to send an entire second copy, which puts the burden of dealing with the mistake on him -- not ideal. Using a patch manager, I could have removed the incorrect file and attached the correct one, adding a followup noting what I did. * If a patch is in the patch manager, it lets people try it out if they want to (or desperately need that specific functionality, and aren't worried about it passing the maintainer's scrutiny). Granted, you can email dillo-dev, but that cuts into the low-traffic guideline. * Adds documentation that the person did some work on a piece of software, which always feels good. That patch entry will probably be around (though closed) for as long as sourceforge is around. * It's easy to see what patches were added for the last release. My table rendering patch made it into CVS (Woohoo! Thanks Jorge!), but there isn't a Changelog entry or any other way of knowing that the table rendering code differs in this release. If all of a sudden a webpage starts crashing CVS dillo, and the only changes since the last working copy were two config file parsing changes and some rendering change it's handy to have a document that you can glance at and say "ah, must be the XYZ changes...I'll start looking there for the source of the problem". * Security. Not all patch managers do this, but SF's allows you to log in and submit authenticated patches. Maybe not a huge deal for dillo since dillo isn't a server, but it avoids spoofed email "from core developers" to Jorge that could introduce unwanted "functionality". * Reliability. Right now, pending patches are spread out around the world on various hard drives (or may even have been deleted). Jorge probably has a complete set, but what if he suffers a hard drive crash? If SF has a set of patches, suddenly having everything vanish in one catastrophic incident isn't an issue. * Finally, with a patch manager, you know that the patch got there. There's no "was my mailserver misbehaving, or is the maintainer just busy ATM?" worries. Cons: * Dillo doesn't currently do https, which you have to do if you want patches attributed to a SF account and authenticated. Dunno whether you can submit non-logged in patches with dillo -- I haven't tried. (Cookie support may even be required...again, I haven't looked) I can see that it might be a bit embarassing to currently have to use a browser other than dillo to submit patches for dillo. Anyway, is there some reason I'm missing? My own currently pending dillo patches are up at , if anyone wants to try 'em or send feedback. -- Best of luck, Mark Schreiber Re: [Dillo-dev]CVS From: Jorgen Viksell - 2002-03-23 02:39 Hi! fre 2002-03-22 klockan 08.43 skrev Calvin Arndt: > downloaded CVS yesterday... says 0.6.5pre1!!!!!!!!! >=20 > compiled -- successfully >=20 > executed -- successfully >=20 > tried http://www.google.com >=20 > tried to set my preferences on google.com >=20 > :-(=20 >=20 > google say's cookies aren't enabled (even though they are in dillorc) >=20 > google.com ACCEPT You mention dillorc, so just to be sure, That line should be in ~/.dillo/cookiesrc Also, the line should probably be one of: .google.com ACCEPT http://www.google.com ACCEPT HTH, J=F6rgen Re: [Dillo-dev]download behavior From: Imad Hussain - 2002-03-22 13:03 On Fri, 22 Mar 2002 13:24:45 +0100 Tomas wrote: > I just wondering if this behaviour is what you want or is going to be > change in the future. > > Also ask if there is any planning of implement resume for file > downloading. Y'know, it would be pretty nice if you could force dillo to use an external utility (e.g. wget) to take care of file downloads completely. I'd imagine the changes necessary to the source would not be monumental in scope -- and for those without wget (or whatever), this could be controlled by an option in .dillorc. It takes much of the burden off of dillo itself; after all, why reinvent the wheel? -- Best, Imad Hussain +========== GBAfan Editor in Chief == http://www.gbafan.net ==========+ +===== OpenBSD. Functional. Secure. Free. http://www.openbsd.org =====+ [Dillo-dev]download behavior From: Tomas - 2002-03-22 12:24 Attachments: Message as HTML Hi all, I'm trying to do a downloading button in the toolbar like it's said in the documentation just to understand dillo source. Now the files are cached so if you donwload a couple of big files the amount of memory used by dillo grows in the same size.=20 Also if you try to save a very big file like an iso-image of 600Mb it fails to allocate this amount of memory:=20 GLib-ERROR **: could not allocate 667461632 bytes aborting... Aborted I just wondering if this behaviour is what you want or is going to be change in the future. Also ask if there is any planning of implement resume for file downloading. greets=20 Tomas [Dillo-dev]CVS From: Calvin Arndt - 2002-03-22 07:43 downloaded CVS yesterday... says 0.6.5pre1!!!!!!!!! compiled -- successfully executed -- successfully tried http://www.google.com tried to set my preferences on google.com :-( google say's cookies aren't enabled (even though they are in dillorc) google.com ACCEPT set number of results=100 saved prefs no of results returned == 10 :-( :-( cookies are not detected by google! just thought you might like to know... cal... __________________________________________________ Do You Yahoo!? Yahoo! Movies - coverage of the 74th Academy Awards® http://movies.yahoo.com/ Re: [Dillo-dev]navigation feature From: John Utz - 2002-03-22 00:06 i sent a patch in months ago to do the 'back' part. but it never showed up in the release (dillo-6.3). sort of takes the wind out of one's sails when it comes to contributing patches..... On Thu, 21 Mar 2002, Imad Hussain wrote: > On Thu, 21 Mar 2002 00:04:25 +0200 (GMT-2) > madis@cy... wrote: > > > > > here is a small patch to add back/forward commands as first items into > > dillo context menu. why? because i got quite used to opera mouse > > gestures and after that found in galeon, that this kind of context > > menu usage is quite comfortable, when compared with moveing mouse over > > the screen to back button... and then back to scroll... i have used > > this patch now some time and it seems to be good for myself, so maybe > > its useful? > > Oh, I'd been looking for a similar patch -- glad to see it's so simple. > ;) Didn't Dillo used to have back and forward on the context menu? Or > was that something just in the literature a few versions back? > > > -- John L. Utz III john@ut... Idiocy is the Impulse Function in the Convolution of Life Re: [Dillo-dev]navigation feature From: Imad Hussain - 2002-03-21 23:51 On Thu, 21 Mar 2002 00:04:25 +0200 (GMT-2) madis@cy... wrote: > > here is a small patch to add back/forward commands as first items into > dillo context menu. why? because i got quite used to opera mouse > gestures and after that found in galeon, that this kind of context > menu usage is quite comfortable, when compared with moveing mouse over > the screen to back button... and then back to scroll... i have used > this patch now some time and it seems to be good for myself, so maybe > its useful? Oh, I'd been looking for a similar patch -- glad to see it's so simple. ;) Didn't Dillo used to have back and forward on the context menu? Or was that something just in the literature a few versions back? -- Best, Imad Hussain +========== GBAfan Editor in Chief == http://www.gbafan.net ==========+ +===== OpenBSD. Functional. Secure. Free. http://www.openbsd.org =====+ [Dillo-dev]plugin interface redesign? From: Geoff Lane - 2002-03-21 15:34 How is the new plugin interface coming along? I have some ideas about pop3:, imap: and news: I'd like to try which really should be implemented as plugins but have been reluctant to start without at least a hint of what the new interface will look like. Thanks -- /\ Geoff. Lane. /\ Manchester Computing /\ Manchester /\ M13 9PL /\ England /\ Re: [Dillo-dev]About bug #81 From: Andreas Schweitzer - 2002-03-21 14:44 ???? What the ... did my mail client just do ? I wanted to suggest solution 3 in combination with Bug 245 (broken images). And one could also have 2 icons - one for broken image and one for unkmown type. Cheers Andreas (Hoping that bottom quoting makes this mail go through ...) On Thu, Mar 21, 2002 at 09:38:49AM -0500, Andreas Schweitzer wrote: > Hi > > > Ok, that seems quite fine to me, but there is a little problem with > > non-root URL's. For example: > > http://www.libmng.com/MNGsuite/basic_img.html > > > > Has 3 non-root images. What should I do in this case? > > > > _______________________________________________ > Dillo-dev mailing list > Dillo-dev@li... > https://lists.so....net/lists/listinfo/dillo-dev -- Department of Physics & Astronomy and Center for Simulational Physics University of Georgia !!! NEW !!! : Phone ++1 (706) 583 8227 Athens, GA 30602-2451 Fax ++1 (706) 542 2492 USA http://dilbert.physast.uga.edu/~andy/ Re: [Dillo-dev]About bug #81 From: Andreas Schweitzer - 2002-03-21 14:38 Hi > Ok, that seems quite fine to me, but there is a little problem with > non-root URL's. For example: > http://www.libmng.com/MNGsuite/basic_img.html > > Has 3 non-root images. What should I do in this case? > [Dillo-dev]About bug #81 From: Livio Baldini Soares - 2002-03-21 12:38 Hello! I've started fixing bug #8 yesterday (trying to view anything other than text, images or HTML). What I did was to offer the user a `download' option. So far it's working out ok, and I try to click on a link that points to a FLASH file, PS, PDF, etc, it instantly opens up the "Save File as..." window. Ok, that seems quite fine to me, but there is a little problem with non-root URL's. For example: http://www.libmng.com/MNGsuite/basic_img.html Has 3 non-root images. What should I do in this case? [1] Should I make instantly open 3 "Save as..." windows? (not very nice for the user.. imagine 20 MNG images on a single page :(. [2] Should I open the windows one at a time? (ugly solution too...) [3] Should I ignore non-root files? I'm leaning towards solution number 3. So my patch would only change Dillo's behaviour with regard to root URLs, i.e. when the user consciously clicks on a certain link/file which Dillo doesn't treat. But I'm not sure this is the best solution. Anyone have any suggestions? Better ideas are welcome! ;) best regards! -- Livio [Dillo-dev]CIPSGA From: Jorge Arellano Cid - 2002-03-20 22:45 Hi everyone! [there's more to this mail than what's apparent to the eye] On Wed, 20 Mar 2002, Livio Baldini Soares wrote: > > > > Congratulations on this patch J=F6rgen and Lars! > > > > FWIW, I made a 35Kb patch over the original 31Kb original > > version, fixing a big list of problems, and no one noticed? :( > > Of course we notice (at least I do!), Jorge. But you get no thanks, > you're the maintainer! It's your _job_! ;-) ! :-) > Seriously though, I have a feeling that Dillo has less developers > than most open source projects I've seen out there (I could be > wrong). So I have an "evil plan" to stimulate our current developers to > produce more & better (giving them lots of thanks.. and who knows, > beer too! ;). Hope it starts working soon! Pay attention please! There's a lot of background activity here, and one of those things is the move from SF to a new site. Livio has uploaded the site already (facilities were provided by cipsga, thanks go to Marcelo, Kov and Henrique). The new site's at: http://dillo.cipsga.org.br/ It hasn't yet been blessed as our official site, but please go there and test it, specially those of you from Europe, so we can get a feel of how the connection behaves. As for the bugtrack, keep using the SF one, until we decide to keep going from cipsga. The next step is to create the mailing-lists there, and to set search capabilities. Probably CVS will be the last thing to move. I wish to release the next dillo version from cipsga, and thus make it official "urbi et orbi". Most of you ignore the fact that SF download counters had been wrong for more than a year (yes I know what I'm saying); just think of all the statistics, project ratings, press coverage et related stuff that has been silently delivered broken... Well, I hope we can finally get a good estimate of our current user base. From the extensive mail archive I have here, comes the fact that dillo is known in at least 37 different countries!!! (yes, from every continent!) Dillo project is steadily gaining more attention, and I'm sure that the new release will be another big step forward as lots of people will be able to start using their web-based mail with dillo. As for new developers, I've always said that this is not an easy project to work in. The average is 6 months for new developers to start sending clean patches, and now, I'm crossing fingers for some freshmen to surpass the magic threshold. With regard to funding (remember 2002 plans?), I'm working on a presentation page to address meta information as: objectives, current state, future plans, whys, whats, advantages, proyections, required money, etc. In summary, very interesting info, aimed to the non-technical reader but worth reading to all of us. Best Jorge.- RE: [Dillo-dev]Dillo on Solaris 8.. From: Yu-Fong Cho - 2002-03-20 22:11 Please give me some information about this. I know this is not a development problem, but I am trying install Dillo on different platforms for interest. I have done on FreeBSD 4.5 and now are trying on Solaris 8 x86. Just give me some information about this and I'll appreciate it. Thank you. -----Original Message----- From: dillo-dev-admin@li... [mailto:dillo-dev-admin@li...]On Behalf Of Yu-Fong Cho Sent: Tuesday, March 19, 2002 12:46 PM To: Dillo Dev Subject: [Dillo-dev]Dillo on Solaris 8.. Hi, Has anyone tried run Dillo 0.6.4 on Solaris 8 before? I compiled and installed it. It seems be ok except the way it handle jpeg image. The jpeg image shows only black and while mode and has some vertical line on it. It isn't right. I have GNU jpeg and zlib installed. Does anyone know what is going on? I tried Dillo on FreeBSD 4.5 before and it was ok. Appreciate any information. Thank you. Yu-Fong Cho _______________________________________________ Dillo-dev mailing list Dillo-dev@li... https://lists.so....net/lists/listinfo/dillo-dev [Dillo-dev]navigation feature From: - 2002-03-20 21:53 Attachments: patch-dillo-menu here is a small patch to add back/forward commands as first items into dillo context menu. why? because i got quite used to opera mouse gestures and after that found in galeon, that this kind of context menu usage is quite comfortable, when compared with moveing mouse over the screen to back button... and then back to scroll... i have used this patch now some time and it seems to be good for myself, so maybe its useful? -- mzz Re: [Dillo-dev]Cookies patch on Solaris From: Livio Baldini Soares - 2002-03-20 21:22 Howdy! Jorge Arellano Cid writes: > > Hi there, > > > Otherwise, I could use dillo with cookies happily (on AIX with strsep) and > > it only was "vulnerable" to the NULL pointer bug that was fixed > > recently. > > > > > Congratulations on this patch Jörgen and Lars! > > FWIW, I made a 35Kb patch over the original 31Kb original > version, fixing a big list of problems, and no one noticed? :( Of course we notice (at least I do!), Jorge. But you get no thanks, you're the maintainer! It's your _job_! ;-) Seriously though, I have a feeling that Dillo has less developers than most open source projects I've seen out there (I could be wrong). So I have an "evil plan" to stimulate our current developers to produce more & better (giving them lots of thanks.. and who knows, beer too! ;). Hope it starts working soon! BTW: Oh Jorge... congratulations you too on this nice patch! (HEheh.. my evil plan is being extended to maintainers ;) (... sorry 'bout the nonsense e-mail ...) best regards to all! -- Livio Re: [Dillo-dev]Cookies patch on Solaris From: Jorge Arellano Cid - 2002-03-20 20:48 Hi there, > Otherwise, I could use dillo with cookies happily (on AIX with strsep) an= d > it only was "vulnerable" to the NULL pointer bug that was fixed > recently. > > > Congratulations on this patch J=F6rgen and Lars! FWIW, I made a 35Kb patch over the original 31Kb original version, fixing a big list of problems, and no one noticed? :( This is very important to me because that was the main reason for the cookies patch not being integrated earlier; there was a lot to do, fix and stabilize. Not lack of interest! BTW, the bugtrack shows 95%. i.e. there're still things to work on; as the strsep() for instance. Thanks Jorge.- Re: [Dillo-dev]Cookies patch on Solaris From: Jorge Arellano Cid - 2002-03-20 20:48 Hi, On Tue, 19 Mar 2002, Livio Baldini Soares wrote: > Anyway, CVS Dillo works well with Linux, but doesn't compile on > Solaris (5.7), unless I --disable-cookies. The problem is that the new > cookies.c uses the strsep() function, which is not available. I have > substituted the strsep() with strtok() which, if I'm not mistaken, is > a POSIX call, and does the "same" thing, no? > > This is an _untested_ patch: > http://www.ime.usp.br/~livio/dillo/patches/strsep_to_strtok.diff We had get rid of strtok() due to its potential problems, so I decided to implement a custom strsep in misc.c. Now in CVS! Please test it. Cheers Jorge.- Re: [Dillo-dev]Cookies patch on Solaris From: Livio Baldini Soares - 2002-03-20 17:31 Jorgen Viksell writes: > Hi! [...] > Unfortunately, strtok() doesn't handle empty pieces (in this case, two > tabs in a row). Ah.. you're right! :( > Looks like we have to write our own stuff here. Shouldn't take too long > though... Well it seems that someone (?) implemented a simple version of strsep() inside misc.c, and now the Solaris version is working fine once again. best regards, -- Livio Re: [Dillo-dev]Cookies patch on Solaris From: Jorgen Viksell - 2002-03-20 15:48 Hi! ons 2002-03-20 klockan 03.19 skrev Livio Baldini Soares: > Hello! >=20 > Well, I'm happy to see that Dillo CVS now has Cookies :-)=20 > I haven't been able to test them though, but I guess they be very > useful to me when authentication and SSL are implemented. >=20 > Anyway, CVS Dillo works well with Linux, but doesn't compile on > Solaris (5.7), unless I --disable-cookies. The problem is that the new > cookies.c uses the strsep() function, which is not available. I have > substituted the strsep() with strtok() which, if I'm not mistaken, is > a POSIX call, and does the "same" thing, no? Unfortunately, strtok() doesn't handle empty pieces (in this case, two tabs in a row). Looks like we have to write our own stuff here. Shouldn't take too long though... > This is an _untested_ patch: > http://www.ime.usp.br/~livio/dillo/patches/strsep_to_strtok.diff >=20 > Congratulations on this patch J=F6rgen and Lars! Thanks! :-) > best regards! =20 >=20 > -- =20 > Livio J=F6rgen Re: [Dillo-dev]Cookies patch on Solaris From: Andreas Schweitzer - 2002-03-20 06:37 Hi > Anyway, CVS Dillo works well with Linux, but doesn't compile on > Solaris (5.7), unless I --disable-cookies. The problem is that the new > cookies.c uses the strsep() function, which is not available. I have > substituted the strsep() with strtok() which, if I'm not mistaken, is > a POSIX call, and does the "same" thing, no? I did the same thing on AIX. However, AIX has also other issues :-) Otherwise, I could use dillo with cookies happily (on AIX with strsep) an= d it only was "vulnerable" to the NULL pointer bug that was fixed recently. > Congratulations on this patch J=F6rgen and Lars! I second that ! :-) Cheers Andreas --=20 Department of Physics & Astronomy and Center for Simulational Physics University of Georgia !!! NEW !!! : Phone ++1 (706) 583 8227 Athens, GA 30602-2451 Fax ++1 (706) 542 2492 USA http://dilbert.physast.uga.edu/~andy/ [Dillo-dev]Cookies patch on Solaris From: Livio Baldini Soares - 2002-03-20 02:19 Hello! Well, I'm happy to see that Dillo CVS now has Cookies :-) I haven't been able to test them though, but I guess they be very useful to me when authentication and SSL are implemented. Anyway, CVS Dillo works well with Linux, but doesn't compile on Solaris (5.7), unless I --disable-cookies. The problem is that the new cookies.c uses the strsep() function, which is not available. I have substituted the strsep() with strtok() which, if I'm not mistaken, is a POSIX call, and does the "same" thing, no? This is an _untested_ patch: http://www.ime.usp.br/~livio/dillo/patches/strsep_to_strtok.diff Congratulations on this patch Jörgen and Lars! best regards! -- Livio [Dillo-dev]Dillo on Solaris 8.. From: Yu-Fong Cho - 2002-03-19 17:46 Hi, Has anyone tried run Dillo 0.6.4 on Solaris 8 before? I compiled and installed it. It seems be ok except the way it handle jpeg image. The jpeg image shows only black and while mode and has some vertical line on it. It isn't right. I have GNU jpeg and zlib installed. Does anyone know what is going on? I tried Dillo on FreeBSD 4.5 before and it was ok. Appreciate any information. Thank you. Yu-Fong Cho [Dillo-dev]Re : Dillo library and binding From: Christophe Baillon - 2002-03-18 13:48 Thanks for your help. Your code works fine ! I know Dw is immature, but Dillo is the only light browser I have found interesting. Dillo is exactly what i need for my work ! We use Dillo 0.6.4 for the Dillo library and binding. The Functionnalities provided by widgets in 0.6.4 are far above our expectations. Regards >Hi Christophe, >On Fri, Mar 08, 2002 at 10:42:30PM +0000, Christophe Baillon wrote: >> I am working on a Ada 95 binding of Dillo widgets. (In GPL of course!) >> To realize it, I need to isolate widgets in a C library. >> I have almost achieved it. >> >> Libdw contains only dillo widget, without IO and cache (I removed >> Dw_Image at the moment) I kept few important dillo file, >> like url.c, findtext.c ... >> >> The goal is to profit of specific dillo widget to make >> rendering in embedded applications. > >Just a few warnings before: 1. Dw is very immature, and will change >often. 2. Dw contains many ad hoc implementations, partly only >suitable for the code of dillo. E.g., there are problems when adding >other widgets than DwPage's to a table (will change a bit). > >Of course, if you have suggestions or improvements, send them to me. > >> I've done a C test program . It can create new dw_style, >> but there is a "segmentation fault" when using >> a_Dw_gtk_scrolled_window_set_dw to set a dw children (Dw_Ruler for >> instance). >> [...] > >The attached example works. See comments for details (especially: a >widget must always have a style). > >Sebastian Re: [Dillo-dev]crash & efence From: Jorge Arellano Cid - 2002-03-15 14:39 Livio, > Jorge Arellano Cid writes: > > > > As indexed GIFs have at most 256 colors, and a corrupted gif > > can declare less colors (or zero), and try to access out of > > bounds bytes afterwards, the 256 constant provides a kind of > > protection against SEGFAULTS. > > [...] > The bug is this. Suppose an Image has only 5 colors. So you have the > image buffer (which is actually entry->Data, or CacheClient->Buf), > which is allocated in the IO system having a fixed size (say 200). So > we're trying to copy 256 bytes out of a 200 byte long buffer... see > the problem? > > I have a quick fix which was, copy _only_ gif->NumColors from the > image Buf, that way we don't memcpy() from invalid memory, _but also_ > alloc more colors (and set them o zero) so if a problem GIF tries to > access an out of bounds colors we get no segfault. > > Watcha think? OK, I just made a few changes and commited to CVS. Please test it. Regards Jorge.- Re: [Dillo-dev]crash & efence From: Livio Baldini Soares - 2002-03-15 03:32 Hello! Jorge Arellano Cid writes: > > As indexed GIFs have at most 256 colors, and a corrupted gif > can declare less colors (or zero), and try to access out of > bounds bytes afterwards, the 256 constant provides a kind of > protection against SEGFAULTS. > > After a fast revision, I can't find anything wrong with it yet, > but what looks very suspicious to me is the 'version' number that > a_Dicache_set_cmap is receiving. > --i.e. #1 in http://www.zone.ee/myzz/dillo/cache-crash-gdb > > That number comes from a_Dicache_add_entry and it should be > a simple counter (AFAIR). Livio is the author of that code, so > this question is for him! :) > Jorge, I fixed that some time ago. Look here: http://cvs.so....net/cgi-bin/viewcvs.cgi/dillo/dillo/src/dicache.c.diff?r1=1.19&r2=1.20 And that actually doesn't matter at all. It's just a version number (doesn't necessarily have to small, or even positive for the matter). The bug is this. Suppose an Image has only 5 colors. So you have the image buffer (which is actually entry->Data, or CacheClient->Buf), which is allocated in the IO system having a fixed size (say 200). So we're trying to copy 256 bytes out of a 200 byte long buffer... see the problem? I have a quick fix which was, copy _only_ gif->NumColors from the image Buf, that way we don't memcpy() from invalid memory, _but also_ alloc more colors (and set them o zero) so if a problem GIF tries to access an out of bounds colors we get no segfault. Watcha think? The patch is at: http://www.ime.usp.br/~livio/dillo/patches/gif_seg.diff best regards! -- Livio Re: [Dillo-dev]crash & efence From: Jorge Arellano Cid - 2002-03-15 02:01 On Wed, 13 Mar 2002 madis@cy... wrote: > On 11 Mar 2002, Lars Clausen wrote: > > > Yes, we've noticed that NULL bug, and how there doesn't seem to be a way to > > get a NULL cookie in the first place. I'm leaning towards an explanation > > of general memory corruption. Did you try running without the cookies > > patch, but linked with efence? I notice that a_Dicache_set_cmap is called > > only from gif.c, and there with num_colors = 256 rather than > > gif->NumColors. That could cause the memcpy to run over the edge of cmap, > > maybe. > > i tried now dillo without cookies and with efence, and the memcpy bug > didn't go away with cookies ;) anyway, replacing this constant 256 in > gif.c with "gif->NumColors < 256 ? gif->NumColors : 256" seems to > solve this memory corruption problem, but since i don't know exactly why > this constant was put there in first place i'm not sure, whether this is a > good fix (or is this < 256 check needed at all, i put it there because > gifs shouldn't have more colors and suspected that constant 256 was > there because some invalid gif caused NumColors go over 256). > http://www.zone.ee/myzz/dillo/patch-dillo-gif As indexed GIFs have at most 256 colors, and a corrupted gif can declare less colors (or zero), and try to access out of bounds bytes afterwards, the 256 constant provides a kind of protection against SEGFAULTS. After a fast revision, I can't find anything wrong with it yet, but what looks very suspicious to me is the 'version' number that a_Dicache_set_cmap is receiving. --i.e. #1 in http://www.zone.ee/myzz/dillo/cache-crash-gdb That number comes from a_Dicache_add_entry and it should be a simple counter (AFAIR). Livio is the author of that code, so this question is for him! :) > but fixing this gif issue didn't fixed segfault totally - when i went to > http://www.delfi.ee site the efence linked dillo crashed happily again and i > discovered bug in IO.c function IO_callback: when IO_read function decided > to close and delete io, following error check tried to access this deleted > structure and set its Status field to -EIO, which wasn't nice at all. > using IO_get again when error occured and checking whether it returned > NULL solved this. > http://www.zone.ee/myzz/dillo/patch-dillo-IO I reviewed this one too, and you're right. The only thing I need is your name for the Changelog entry! Cheers Jorge.- Re: [Dillo-dev]crash & efence From: Jorge Arellano Cid - 2002-03-14 13:47 Hi there! On 12 Mar 2002, Jorgen Viksell wrote: > Hi, > > tis 2002-03-12 klockan 01.45 skrev madis@cy...: > > > > i use dillo-0.6.4 patched with patch-cookies-8jan (without openssl > > patch) and it crashed occasionally, most often on theregister.co.uk. so= i > > tried to use gdb to find out why, and yes, there was a NULL pointer > > reference in a_Cookies_get function (look http://www.zone.ee/myzz/dillo= /cookie-crash-gdb). > > but after reading code in cookies.c i had no clue, how this pointer cou= ld > > be NULL without segfaulting before [snip] > > I believe that I've found the problem. It's in a_Cookies_set(), where an > old cookie gets replaced. We remove it from the list but we never > re-insert the new start of the list in the hash table. > > There is a new patch up at: > http://w1.211.telia.com/~u21114235/patch-cookies-12mar.gz > Lots of changes made. Just to complete the information, specially for those that got eager for the original patch to be commited: I reviewed the first cookie full-patch, it was 31Kb, and found a big list of problems, so I gave it some hard work and ended up with a 35Kb patch on top of it, and there were still plenty of things to do! Now J=F6rgen has sent me a new patch (on top of both) that'll be reviewed ASAP. See ya' Jorge.- Re: [Dillo-dev]Bug 306 - ">" sign in attribute string From: Jorge Arellano Cid - 2002-03-13 17:21 Andreas, > I found a bug and instered it as bug 306. The good news : I also > have a patch. But it may be tricky, so I send it in as food for thought. > > The problem is that dillo gets confused when the > sign is in an attribute > string (like http://www.xwinman.org). Simply accepting a > inside a string > will break way too many web pages. E.g. "slahsdot->Read more" on any > article has > > Near the login form. Looks like broken HTML to me ... Yeah, that page is broken... But, the http://www.xwinman.org case may be different. Considering the XML SPEC for XHTML, we find: AttValue ::== '"' ([^<&"] | Reference)* '"' | "'" ([^<&'] | Reference)* "'" i.e. a bare '>' is allowed, but not a bare '<'. That's what I did! Please test the CVS and let me know how it went. Cheers Jorge.- [Dillo-dev]Presentation From: Tomas - 2002-03-13 12:05 Attachments: Message as HTML Hi all, I'm spanish (sorry for my english in advance ;) and just finished computer = science, i'm involved in a non-profit project of a free internet workplace in Utrecht (Netherlands) http://www.puscii.nl and i normailly used GNU/Debian = as OS. First of all congratulations to all the people involved in dillo for all this good work, i found needed an alternative graphical web wrobser with lo= w-resources requirements in the free-software scene so i think dillo cover this hole. And is quite usefull now even thought is in alpha state. Now that i have time i will like to contribute to dillo project so i'm subscribed to dillo-dev list and now i'm reading and trying to understa= nd the code of dillo, also reading gtk manual and html specifications as you wrote in the documentation. I've found the code quite clean but i need some time to understand it all, i also need time to be fluence with gdb, CV= S and=20 other tools normailly used (any sugestion is apreciated :) greets Tomas Re: [Dillo-dev]crash & efence From: - 2002-03-13 01:41 On 11 Mar 2002, Lars Clausen wrote: > Yes, we've noticed that NULL bug, and how there doesn't seem to be a way to > get a NULL cookie in the first place. I'm leaning towards an explanation > of general memory corruption. Did you try running without the cookies > patch, but linked with efence? I notice that a_Dicache_set_cmap is called > only from gif.c, and there with num_colors = 256 rather than > gif->NumColors. That could cause the memcpy to run over the edge of cmap, > maybe. i tried now dillo without cookies and with efence, and the memcpy bug didn't go away with cookies ;) anyway, replacing this constant 256 in gif.c with "gif->NumColors < 256 ? gif->NumColors : 256" seems to solve this memory corruption problem, but since i don't know exactly why this constant was put there in first place i'm not sure, whether this is a good fix (or is this < 256 check needed at all, i put it there because gifs shouldn't have more colors and suspected that constant 256 was there because some invalid gif caused NumColors go over 256). http://www.zone.ee/myzz/dillo/patch-dillo-gif but fixing this gif issue didn't fixed segfault totally - when i went to http://www.delfi.ee site the efence linked dillo crashed happily again and i discovered bug in IO.c function IO_callback: when IO_read function decided to close and delete io, following error check tried to access this deleted structure and set its Status field to -EIO, which wasn't nice at all. using IO_get again when error occured and checking whether it returned NULL solved this. http://www.zone.ee/myzz/dillo/patch-dillo-IO i'll try this new cookies patch tomorrow night... currently morning is too close :) [Dillo-dev]Multidimensional Bookmarks From: Bruce Alon King - 2002-03-13 01:26 I've been using Dillo for a couple of months now, resorting to Opera or Mozilla when necessary. I started running it because of low memory; now that I've upgraded my RAM I run it because I want to. So: KUDOS! So: I've always wanted some sort of reading system. Basically, I want a bookmark system that saves state so that I can not only remember the page, but where I was on that page, as well as a history of the pages visited. That way, say I'm browsing through Gtk documentation, I read through several pages, skip around, and then have to go to bed. I can save my bookmark in a .bkmk file (or whatever) and the next morning, bright and early, start exactly where I left off. I can even click the back button, yada yada. I'm trying to find some time to read through the source code and start contributing to this project so I'm perfectly willing to implement it sooner or later. I just wondered, does this idea interest anybody else or is it just me? I've been reading more and more books online (Structure and Interpretation of Computer Programs, Baen Free Library, et al) and I find it increasingly frustrating that I have to spend time navigating to where I once was in order to start reading again. In a paper book, I turn to my bookmark and there I am, within millimeters of the correct location. The other question is: any guesses on how hard this would be to implement? It seems like it should be pretty simple, but I haven't yet had the time to go through any of the dillo source code. -- Bruce Alon King The frogs are dancing and frollicking among the weeds. They smile. Re: [Dillo-dev]crash & efence From: Jorgen Viksell - 2002-03-12 20:35 Hi, tis 2002-03-12 klockan 01.45 skrev madis@cy...: >=20 > i use dillo-0.6.4 patched with patch-cookies-8jan (without openssl=20 > patch) and it crashed occasionally, most often on theregister.co.uk. so i= =20 > tried to use gdb to find out why, and yes, there was a NULL pointer=20 > reference in a_Cookies_get function (look http://www.zone.ee/myzz/dillo/c= ookie-crash-gdb). > but after reading code in cookies.c i had no clue, how this pointer could= =20 > be NULL without segfaulting before [snip] I believe that I've found the problem. It's in a_Cookies_set(), where an old cookie gets replaced. We remove it from the list but we never re-insert the new start of the list in the hash table. There is a new patch up at: http://w1.211.telia.com/~u21114235/patch-cookies-12mar.gz Lots of changes made. Cheers, J=F6rgen Re: [Dillo-dev]Dillo on Debian/Hurd - Compiled (see below) but malfunctions... From: Sam Dennis - 2002-03-12 19:40 Some time around 7 o'clock PM on March 12, a Tuesday, Sammy Mannaert wrote: > "B. Douglas Hilton" wrote: > > > > > > It cheerfully complied with no problems, loads up the default > > window in X, and then kind of goes out to lunch if you try > > to open a file or url. > > you can also compile dillo by changing defines in src/dns.c > (outcommenting #define G_DNS_THREADED) > this way dillo will not use multithreading :) Unfortunately, asynchronous DNS is not the only thing dillo uses threads for. IO/file.c contains other code, and no alternative method. -- r@,,+2 'a,kd" Re: [Dillo-dev]Dillo on Debian/Hurd - Compiled (see below) but malfunctions... From: Sammy Mannaert - 2002-03-12 18:21 "B. Douglas Hilton" wrote: > > > It cheerfully complied with no problems, loads up the default > window in X, and then kind of goes out to lunch if you try > to open a file or url. you can also compile dillo by changing defines in src/dns.c (outcommenting #define G_DNS_THREADED) this way dillo will not use multithreading :) Re: [Dillo-dev]Dillo library, and binding From: Sebastian Geerken - 2002-03-12 18:11 Attachments: test_dw.c Hi Christophe, On Fri, Mar 08, 2002 at 10:42:30PM +0000, Christophe Baillon wrote: > I am working on a Ada 95 binding of Dillo widgets. (In GPL of course!) > To realize it, I need to isolate widgets in a C library. > I have almost achieved it. > > Libdw contains only dillo widget, without IO and cache (I removed > Dw_Image at the moment) I kept few important dillo file, > like url.c, findtext.c ... > > The goal is to profit of specific dillo widget to make > rendering in embedded applications. Just a few warnings before: 1. Dw is very immature, and will change often. 2. Dw contains many ad hoc implementations, partly only suitable for the code of dillo. E.g., there are problems when adding other widgets than DwPage's to a table (will change a bit). Of course, if you have suggestions or improvements, send them to me. > I've done a C test program . It can create new dw_style, > but there is a "segmentation fault" when using > a_Dw_gtk_scrolled_window_set_dw to set a dw children (Dw_Ruler for > instance). > [...] The attached example works. See comments for details (especially: a widget must always have a style). Sebastian [Dillo-dev]Dillo on Debian/Hurd - Compiled (see below) but malfunctions... From: B. Douglas Hilton - 2002-03-12 01:02 Just thought I'd post a heads up on what I did. A few weeks ago I made an experimental patch to GNU Posix Threads (Pth), which provides a non-preemptive pthread compatibility layer, so that it would compile on the GNU Hurd OS. Hurd only has c-threads built into the gnumach microkernel, and several high-powered Debian developers are working on making the ultimate pthreads implementation, which will likely be a while before its ready :-) So, hacker that I am, I made some really dangerous hacks to pth and got it to compile (not saying it works right), and used that to build Dillo on Hurd. It cheerfully complied with no problems, loads up the default window in X, and then kind of goes out to lunch if you try to open a file or url. Still, I thought it was pretty cool. Like the browser, it is awesome... right on with the nix philosophy: make a small powerfull tool to do exactly one thing, and then make it ubiquitous because its so small and fast! Later! - Doug Re: [Dillo-dev]crash & efence From: Lars Clausen - 2002-03-12 00:45 On Tue, 12 Mar 2002, madis@cy... wrote: >=20 > i use dillo-0.6.4 patched with patch-cookies-8jan (without openssl patch) > and it crashed occasionally, most often on theregister.co.uk. so i tried > to use gdb to find out why, and yes, there was a NULL pointer reference > in a_Cookies_get function (look > http://www.zone.ee/myzz/dillo/cookie-crash-gdb). but after reading code > in cookies.c i had no clue, how this pointer could be NULL without > segfaulting before and tried linking with electric fence memory debugger > using -lefence. when linked with efence dillo segfaulted on all web pages > with gif images in a_Dicache_set_cmap function > (http://www.zone.ee/myzz/dillo/cache-crash-gdb). sorrily i had not enough > understanding about dicache to debug it... but seems that there is some > bug. Yes, we've noticed that NULL bug, and how there doesn't seem to be a way to get a NULL cookie in the first place. I'm leaning towards an explanation of general memory corruption. Did you try running without the cookies patch, but linked with efence? I notice that a_Dicache_set_cmap is called only from gif.c, and there with num_colors =3D 256 rather than gif->NumColors. That could cause the memcpy to run over the edge of cmap, maybe.=20 -Lars --=20 Lars Clausen (http://shasta.cs.uiuc.edu/~lrclause)| H=E5rdgrim of Numenor "I do not agree with a word that you say, but I |------------------------= ---- will defend to the death your right to say it." | Where are we going, and --Evelyn Beatrice Hall paraphrasing Voltaire | what's with the handbas= ket? [Dillo-dev]crash & efence From: - 2002-03-12 00:35 i use dillo-0.6.4 patched with patch-cookies-8jan (without openssl patch) and it crashed occasionally, most often on theregister.co.uk. so i tried to use gdb to find out why, and yes, there was a NULL pointer reference in a_Cookies_get function (look http://www.zone.ee/myzz/dillo/cookie-crash-gdb). but after reading code in cookies.c i had no clue, how this pointer could be NULL without segfaulting before and tried linking with electric fence memory debugger using -lefence. when linked with efence dillo segfaulted on all web pages with gif images in a_Dicache_set_cmap function (http://www.zone.ee/myzz/dillo/cache-crash-gdb). sorrily i had not enough understanding about dicache to debug it... but seems that there is some bug. Re: [Dillo-dev]' '; or '  ': Comforming to standart From: Eduardo Marcel - 2002-03-11 11:32 I guess we should support the standard. Adopting this kind of workaround = on such=20 common mistakes would only make the mistakes more common. IMHO It's the o= ther browsers' mistake. Dillo is correct :) =20 Regards, Eduardo. On Mon, 11 Mar 2002 16:59:06 +0600 "Nikita V. Borodikhin" wrote: > Hello, All ! >=20 > Should Dillo follow to official standart, where non-breakable > space is ' ' only or it should follow de-facto standart, > where such a space could be written as ' ' + any non-letter > chacter, e.g. a ' ' sharacter ? >=20 > That non-official form of nonbreakable space is succesfully eaten > by IE, NS, Opera, links, lynx and may be many others and is widely > used in the WEB. >=20 > =3D=3D=3D > Nikita V. Borodikhin >=20 > _______________________________________________ > Dillo-dev mailing list > Dillo-dev@li... > https://lists.so....net/lists/listinfo/dillo-dev >=20 --=20 Eduardo Marcel Ma=E7an Gerente de Redes / Network Manager macan@co... Col=E9gio Bandeirantes Re: [Dillo-dev]' '; or '  ': Comforming to standart From: Grigory Bakunov - 2002-03-11 11:18 Date |Mon, 11 Mar 2002 16:59:06 +0600 From |"Nikita V. Borodikhin" Hello! NVB> Hello, All ! NVB> Should Dillo follow to official standart, where non-breakable NVB> space is ' ' only or it should follow de-facto standart, NVB> where such a space could be written as ' ' + any non-letter NVB> chacter, e.g. a ' ' sharacter ? NVB> That non-official form of nonbreakable space is succesfully eaten NVB> by IE, NS, Opera, links, lynx and may be many others and is widely NVB> used in the WEB. My vote to   Look to w3c standarts. It's only one source of power ;) ------------------------------------------------------- Grigory Bakunov ASPLinux Support Team http://www.asplinux.ru [Dillo-dev]' '; or '  ': Comforming to standart From: Nikita V. Borodikhin - 2002-03-11 11:07 Hello, All ! Should Dillo follow to official standart, where non-breakable space is ' ' only or it should follow de-facto standart, where such a space could be written as ' ' + any non-letter chacter, e.g. a ' ' sharacter ? That non-official form of nonbreakable space is succesfully eaten by IE, NS, Opera, links, lynx and may be many others and is widely used in the WEB. === Nikita V. Borodikhin [Dillo-dev]Non iso-8859 rendering/pango From: Mark Schreiber - 2002-03-10 17:46 After looking at bug #299 for an hour or two ("dillo can't render iso entities > 255"), I suspect that the easiest thing to do may be to support pango, which would provide Unicode support, but would entail a *lot* of work and rewriting, both in layout and in internal storage of string data. A great deal of logic would have to be entirely different, at least to do things properly. AFAIK, pango 1.0 is out (ftp://ftp.gtk.org/pub/gtk/v2.0), but not widely used yet. Is anyone currently doing pango support? If not, I'm thinking about maybe making a few small interim patches to handle 8859 approximations of non-iso-8859 characters, like double dashes for an mdash and straight quotes for printers' (aka curly) quotes. -- Best of luck, Mark Schreiber [Dillo-dev]Dillo library, and binding From: Christophe Baillon - 2002-03-08 21:44 Hi ! First, thanks to dillo developpers for this fast browser ! I am working on a Ada 95 binding of Dillo widgets. (In GPL of course!) To realize it, I need to isolate widgets in a C library. I have almost achieved it. Libdw contains only dillo widget, without IO and cache (I removed Dw_Image at the moment) I kept few important dillo file, like url.c, findtext.c ... The goal is to profit of specific dillo widget to make rendering in embedded applications. I've done a C test program . It can create new dw_style, but there is a "segmentation fault" when using a_Dw_gtk_scrolled_window_set_dw to set a dw children (Dw_Ruler for instance). What is wrong in my code ? I think everything is initialized ? Is there special things to do before adding a Dillo Widget to a container ? gdb said it fails in a_Dw_widget_realize, at : if (widget->parent =3D=3D NULL && widget->style->background_color !=3D NULL= ) Code : #include "dw.h" #include "dw_widget.h" #include "dw_gtk_scrolled_window.h" #include "gtk/gtkscrolledwindow.h" // Compiler avec : // gcc -L../lib test.c -g -I ../src `gtk-config --cflags` `gtk-config // --libs` -ldw -o test=20 main(int argc, char **argv) { =20 GtkWidget *fenetre; GtkWidget *label; =20 DwWidget *ma_table; DwWidget *ma_page; DwWidget *ma_ruler; GtkWidget *mon_scroll; DwStyle style_attrs, *mon_style; DwStyleFont font_attrs; =20 gtk_init(&argc, &argv); gdk_rgb_init(); a_Dw_init(); =20 =20 fenetre =3D gtk_window_new (GTK_WINDOW_TOPLEVEL); label =3D gtk_label_new ("Bonjour, ceci est un test"); ma_table =3D a_Dw_table_new(); ma_page =3D a_Dw_page_new(); ma_ruler =3D a_Dw_hruler_new(); mon_scroll =3D a_Dw_gtk_scrolled_window_new(); gtk_scrolled_window_set_policy(GTK_SCROLLED_WINDOW(mon_scroll), GTK_POLICY_AUTOMATIC, GTK_POLICY_AUTOMATIC); gtk_container_add (fenetre, mon_scroll); gtk_widget_show_all(fenetre); a_Dw_gtk_scrolled_window_set_dw (GTK_DW_SCROLLED_WINDOW(mon_scroll), ma_ruler); // Here is segmentation fault=20 =20 =20 /////////////////////////////////// // Initialisation du Dw_Style // ///////////////////////////////// =20 /* Create the font and attribute for the page. */ font_attrs.name =3D "courier"; font_attrs.size =3D 14; font_attrs.bold =3D FALSE; font_attrs.italic =3D FALSE; =20 a_Dw_style_init_values (&style_attrs, fenetre->window); style_attrs.font =3D a_Dw_style_font_new (&font_attrs); style_attrs.color =3D a_Dw_style_color_new (0x0000ff, fenetre->window); mon_style =3D a_Dw_style_new (&style_attrs, fenetre->window); //a_Dw_widget_set_style (plain->dw, plain->style); =20 ///////////////////////////////////////////////////////////// =20 =20 //a_Dw_page_add_widget (ma_page, // ma_ruler, // mon_style); =20 // If i uncomment it, =3D> segmentation fault =20 gtk_main(); } Is anybody make similar works ? Thanks for your help ! Regards --=20 ********************************* *** Christophe Baillon *** *** ITRIS *** *** Informatique Temps R=E9el *** *** Informatique Syst=E8me *** ********************************* Re: [Dillo-dev]dillo (0.6.4) not showing background images ? From: John Utz - 2002-03-08 19:21 for starters, it's a gif! so it aint gonna get displayed unless you add gif functionality to your copy of dillo.... On Fri, 8 Mar 2002, Peter Schb wrote: > hi, >=20 > i=B4m using dillo 0.6.4 on SUSE linux 7.3 (kernel 2.4.10). > dillo is *not* showing background images ! (other images work fine) >=20 > ... >=20 > is this still a missing feature or is something wrong with my configura= tion ?? >=20 > please respond to mailto:schb@gm... >=20 > thanks > peter >=20 >=20 --=20 John L. Utz III john@ut... Idiocy is the Impulse Function in the Convolution of Life [Dillo-dev]dillo (0.6.4) not showing background images ? From: Peter Schb - 2002-03-08 17:59 Attachments: Message as HTML hi, i=B4m using dillo 0.6.4 on SUSE linux 7.3 (kernel 2.4.10). dillo is *not* showing background images ! (other images work fine) ... is this still a missing feature or is something wrong with my = configuration ?? please respond to mailto:schb@gm... thanks peter Re: [Dillo-dev]bug track review From: Sebastian Geerken - 2002-03-08 17:36 On Fri, Mar 01, 2002 at 11:50:46AM -0800, Eric GAUDET wrote: > -- En reponse de "[Dillo-dev]bug track review" de Sebastian Geerken, le > >> 202 > >> Impact: soft error > >> At: run time > >> Type: render bug > >> Comments: When an image is loading over a slow connection, and you > >> start scrolling up and down while it draws, some parts of it may not > >> get updated. HTReproduce: Any big image will do. For instance, try: > >> http://www.photo.net/photo/pcd1765/walk-away-44.tcl > >> WorkedBy: nobody > > > > This is a problem of Dw_widget_queue_draw_area, which would quite hard > > to solve. If somebody is *really* interested in solving this, just ask > > me for a more detailed explanation. An exact explanation of the problem: Dw widgets should normally use the function Dw_widget_queue_draw_area and not draw directly, this is a bit more efficient. This pattern was taken from Gtk+. The implementation of Dw_widget_queue_draw_area is rather simple: It just passes the transformed area to gtk_widget_queue_draw. The problem is, that Gtk+ deals with the coordinates within the Gtk+ widget, what are for Dw viewport coordinates (see doc/Dw.txt for the definitions). What has actually to be queued, are world coordinates, but when the viewport position changes between queueing and actual drawing, we get wrong world coordinates. As in this drawing: When queueing: +----------------+<-- viewport | +----------+ | | | |<-+--- image | | | | | | | | | +----------+ | | | <---------- area to be redrawn | +----------+ | | | | | | | +----------------+ When drawing, after viewport position has changed: +----------------+ | | | | | +----------+ | | | | | | +----------+ | | | <----------- area which is redrawn | +----------+ | | +----------+ | | | <----------- this area should be redrawn | +----------+ | +----------------+ The best fix would of course to re-implement the whole queueing mechanism, this is what I meant with "hard to solve". > My first guess would be to add an int y_painted in the DwImage, to keep track > of what?s already painted, and use that as an origin and update it in > Dw_widget_queue_draw_area. > Another possibility would be toto add a boolean that would be set when > scrolled, and when an image is finished loading invoke a_Dw_widget_queue_draw ? > (Or, I don?t like it so much, but no boolean and force a redraw for when images > finished downloading, perhaps after testing if they are visible) > ... or did I miss something? This is no image specific bug, but IIRC, DwImage is the only widget which calls Dw_widget_queue_draw, the others call Dw_widget_queue_resize, which implies a redraw, but works different. > >> 289 > >> Impact: soft error > >> At: run time > >> Type: render bug > >> Comments: when an image tooltip is display and you scroll the page > >> with the keyboard, dillo fails to render the new scrolled content. > >> HTReproduce: Go to dillo home page, move the mouse cursor on the top > >> image, when the tooltip is displayed, press PgDown: the page scrolls > >> but the bottom is empty. > >> WorkedBy: nobody > > > > This is a Gtk+ bug, it can be simply be reproduced (combination of > > tooltips and a GtkLayout, in which is painted). > > > > Ok. Is it possible to find a workaround ? Like remove the tooltip as soon as a > key even is received. That would also solve the sticky tooltips ?feature?. Or fix it in Gtk+ :-) The sticky tooltips are probably a dillo bug, in the Dw event system, probably some events are still missing. Sebastian Re: [Dillo-dev]bug track review From: Sebastian Geerken - 2002-03-08 17:04 On Fri, Mar 01, 2002 at 03:11:28PM -0600, Lars Clausen wrote: > On Fri, 1 Mar 2002, Sebastian Geerken wrote: > >> 233 Impact: soft error At: run time Type: render bug Comments: dillo > >> doesn't render side-by-side tables (with align=left) correctly. Dillo > >> renders one table per line, no matter what. It seems to be a standard to > >> render such tables on the same line. HTReproduce: > >> http://www.rti-zone.org/dillo/Html.testsuite/table-side.html > >> http://www.dpreview.com/ WorkedBy: nobody > > > > Such a behavior would be actually not quite standard compliant, see > > http://www.w3.org/TR/1999/REC-html401-19991224/struct/tables.html#edef-TABLE: > > > > align = left|center|right [CI] > > Deprecated. This attribute specifies the position of the table > > with respect to the document. Permitted values: > > * left: The table is to the left of the document. > > * center: The table is to the center of the document. > > * right: The table is to the right of the document. > > > > I'm not sure whether we should follow the standard, or common > > use. Anyway, floats will be used for some other purposes. > > Since this is part of an older standard, and in very frequent use, I think > we should support it. And since this kind of floating is (I believe) part > of CSS, we will need it then anyway. This is probably the most common > layout fault I find when using Dillo, I'd *love* to see it fixed. I even > looked into how to do it, but couldn't figure out how to change the > margins. http://www.w3.org/TR/html40/present/graphics.html#h-15.1.3 has > a little more info on floating. Yes, floats will be supported in the future, this is indeed a bit tricky (I also made some thoughts on it). The implementation will use a "float" member in the DwStyle (names should already be similar to attribute names in CSS), so this is an issue which is simple to change in the HTML parser (using "float" or "text_align"). Sebastian [Dillo-dev]Thanks, good work and good luck. From: Stephen Young - 2002-03-08 04:13 To all developers of dillo, Good work. I can't even remember the first time that I used dillo(must of been a one year-two years ago). Each release that i've tried has been getting better and better. Up until now however using dillo wasn't really all that realistic for most of my browsing. However, with this release, it has been quite pleasant with its stability and speed. I usually use Opera/Netscape combination but now I'll be using Dillo/Netscape. Thanks, Stephen Young [Dillo-dev]Dillo support From: Pradeep - 2002-03-07 11:48 Hi , I am trying to port Dillo to a custom board , does Dillo support static gray visuals , waiting for your reply. Thanks Pradeep Re: [Dillo-dev]Small make install issue From: Michael Duelli - 2002-03-04 21:11 D'oh!! I forgot to include the relevant part of the output :-) make install says: ... make[2]: Entering directory `/usr/src/packages/SOURCES/DEV/dillo' make[2]: Nothing to be done for `install-exec-am'. if [ -d /usr/local/etc ]; then \ /usr/bin/ginstall -c -m 644 --backup=t ./dillorc /usr/local/etc/; \ elif [ -d /etc/ ]; then \ /usr/bin/ginstall -c -m 644 --backup=t ./dillorc /etc/; \ fi /usr/bin/ginstall: option `--backup' doesn't allow an argument Try `/usr/bin/ginstall --help' for more information. make[2]: [install-data-local] Error 1 (ignored) make[2]: Leaving directory `/usr/src/packages/SOURCES/DEV/dillo' make[1]: Leaving directory `/usr/src/packages/SOURCES/DEV/dillo' And I don't know why but ginstall --backup gets a t value. ginstall help tells me something about version control and the t value, but not with the --backup flag. BTW, I only do: ./configure make make install I don't specify any special directories or something like that. Of course, it's not a real error. dillo installs and runs just fine. I just wanted to tell you. TIA, Michael ___________________________________ Michael Duelli m.duelli@we... http://glchess.s...net - A 3D chess interface Re: [Dillo-dev]Small make install issue From: Livio Baldini Soares - 2002-03-04 18:53 Hello Michael! Michael Duelli writes: > Hi all, > > when installing dillo as root using ginstall from install (GNU fileutils) > 4.0 I get this tiny error message Hum... where is the error? Isn't this the usual output for 'make install' (or did you accidently do 'make -n install')? Or am I missing something? Does it get to installing the dillorc file at all? > --------------snip---------------- > make[1]: Entering directory `/usr/src/packages/SOURCES/DEV/dillo' > make[2]: Entering directory `/usr/src/packages/SOURCES/DEV/dillo' > make[2]: Nothing to be done for `install-exec-am'. > if [ -d /usr/local/etc ]; then \ > /usr/bin/ginstall -c -m 644 --backup=t ./dillorc /usr/local/etc/; \ > elif [ -d /etc/ ]; then \ > /usr/bin/ginstall -c -m 644 --backup=t ./dillorc /etc/; \ > fi > make[2]: Leaving directory `/usr/src/packages/SOURCES/DEV/dillo' > make[1]: Leaving directory `/usr/src/packages/SOURCES/DEV/dillo' > --------------snap---------------- > > It seems like you have to export the argument for backup as > SIMPLE_BACKUP_SUFFIX. You have to export SIMPLE_BACKUP_SUFFIX _only_ if you'd like to change the default backup suffix (the default is '~'). > Maybe this is different with newer versions of ginstall. Err.. I have access to three diferent machines using three diferent fileutils (3.6 on Solaris, 4.0l on Debian[Potato]/Linux, and 4.1-9 on Debian[Woody]/Linux), and they all work :( Please try to check if the dillorc gets copied. I would like to help you out, but I need to understand what's wrong and how to reproduce your error. best regards, -- Livio Re: [Dillo-dev]Solaris & dillo From: Livio Baldini Soares - 2002-03-04 18:25 Jorge Arellano Cid writes: > > Hi there! Hello Jorge! > Some of the autconf changes (in CVS) broke Solaris detection. > Now there's new code to fix this. It has been partly tested, but > there's a need for a run-test. Someone with a SunOS box please > test it and report back how it goes. # uname -a SunOS jaca 5.7 Generic_106541-04 sun4u sparc Everything seems to be just fine from here. No problems at all! Pthreads are now working again. best regards! -- Livio [Dillo-dev]Solaris & dillo From: Jorge Arellano Cid - 2002-03-04 17:08 Hi there! Some of the autconf changes (in CVS) broke Solaris detection. Now there's new code to fix this. It has been partly tested, but there's a need for a run-test. Someone with a SunOS box please test it and report back how it goes. Thanks Jorge.- [Dillo-dev]Bug 306 - ">" sign in attribute string From: Andreas Schweitzer - 2002-03-04 04:46 Attachments: html.c.diff Hi, I found a bug and instered it as bug 306. The good news : I also have a patch. But it may be tricky, so I send it in as food for thought. The problem is that dillo gets confused when the > sign is in an attribute string (like http://www.xwinman.org). Simply accepting a > inside a string will break way too many web pages. E.g. "slahsdot->Read more" on any article has Near the login form. Looks like broken HTML to me ... Anyways, I attach a patch which looks ahead to determine the order of ", > and < What do you think ? Cheers, Andreas -- Department of Physics & Astronomy and Center for Simulational Physics University of Georgia !!! NEW !!! : Phone ++1 (706) 583 8227 Athens, GA 30602-2451 Fax ++1 (706) 542 2492 USA http://dilbert.physast.uga.edu/~andy/ Re: [Dillo-dev]bug track review From: Lars Clausen - 2002-03-01 21:11 On Fri, 1 Mar 2002, Sebastian Geerken wrote: > Hi, >=20 > I just took some time to take a look on the bug tracking engine. Here > are some comments on a few. [...] >> 233 Impact: soft error At: run time Type: render bug Comments: dillo >> doesn't render side-by-side tables (with align=3Dleft) correctly. Dill= o >> renders one table per line, no matter what. It seems to be a standard = to >> render such tables on the same line. HTReproduce: >> http://www.rti-zone.org/dillo/Html.testsuite/table-side.html >> http://www.dpreview.com/ WorkedBy: nobody >=20 > Such a behavior would be actually not quite standard compliant, see > http://www.w3.org/TR/1999/REC-html401-19991224/struct/tables.html#edef-= TABLE: >=20 > align =3D left|center|right [CI] > Deprecated. This attribute specifies the position of the table > with respect to the document. Permitted values: > * left: The table is to the left of the document. > * center: The table is to the center of the document. > * right: The table is to the right of the document. >=20 > I'm not sure whether we should follow the standard, or common > use. Anyway, floats will be used for some other purposes. Since this is part of an older standard, and in very frequent use, I thin= k we should support it. And since this kind of floating is (I believe) par= t of CSS, we will need it then anyway. This is probably the most common layout fault I find when using Dillo, I'd *love* to see it fixed. I even looked into how to do it, but couldn't figure out how to change the margins. http://www.w3.org/TR/html40/present/graphics.html#h-15.1.3 has a little more info on floating. -Lars --=20 Lars Clausen (http://shasta.cs.uiuc.edu/~lrclause)| H=E5rdgrim of Numenor "I do not agree with a word that you say, but I |----------------------= ------ will defend to the death your right to say it." | Where are we going, a= nd --Evelyn Beatrice Hall paraphrasing Voltaire | what's with the handb= asket? RE: [Dillo-dev]bug track review From: Eric GAUDET - 2002-03-01 19:52 -- En reponse de "[Dillo-dev]bug track review" de Sebastian Geerken, le 01-Mar-2002 : > Hi, >=20 > I just took some time to take a look on the bug tracking engine. Here > are some comments on a few. >=20 >> 202 >> Impact: soft error >> At: run time >> Type: render bug >> Comments: When an image is loading over a slow connection, and you >> start scrolling up and down while it draws, some parts of it may not >> get updated. HTReproduce: Any big image will do. For instance, try: >> http://www.photo.net/photo/pcd1765/walk-away-44.tcl >> WorkedBy: nobody >=20 > This is a problem of Dw_widget_queue_draw_area, which would quite hard > to solve. If somebody is *really* interested in solving this, just ask > me for a more detailed explanation. >=20 My first guess would be to add an int y_painted in the DwImage, to keep t= rack of what=B4s already painted, and use that as an origin and update it in Dw_widget_queue_draw_area.=20 Another possibility would be toto add a boolean that would be set when scrolled, and when an image is finished loading invoke a_Dw_widget_queue_= draw ? (Or, I don=B4t like it so much, but no boolean and force a redraw for whe= n images finished downloading, perhaps after testing if they are visible) ... or did I miss something? >> 233 >> Impact: soft error >> At: run time >> Type: render bug >> Comments: dillo doesn't render side-by-side tables (with align=3Dleft) >> correctly. Dillo renders one table per line, no matter what. It seems >> to be a standard to render such tables on the same line. >> HTReproduce: http://www.rti-zone.org/dillo/Html.testsuite/table-side.h= tml >> http://www.dpreview.com/ >> WorkedBy: nobody >=20 > Such a behavior would be actually not quite standard compliant, see > http://www.w3.org/TR/1999/REC-html401-19991224/struct/tables.html#edef-= TABLE: >=20 > align =3D left|center|right [CI] > Deprecated. This attribute specifies the position of the table > with respect to the document. Permitted values: > * left: The table is to the left of the document. > * center: The table is to the center of the document. > * right: The table is to the right of the document. >=20 > I'm not sure whether we should follow the standard, or common > use. Anyway, floats will be used for some other purposes. >=20 I=B4m affraid I see that in a lot of sites. I=B4d prefer we support it. >> 289 >> Impact: soft error >> At: run time >> Type: render bug >> Comments: when an image tooltip is display and you scroll the page >> with the keyboard, dillo fails to render the new scrolled content. >> HTReproduce: Go to dillo home page, move the mouse cursor on the top >> image, when the tooltip is displayed, press PgDown: the page scrolls >> but the bottom is empty. >> WorkedBy: nobody >=20 > This is a Gtk+ bug, it can be simply be reproduced (combination of > tooltips and a GtkLayout, in which is painted). >=20 Ok. Is it possible to find a workaround ? Like remove the tooltip as soon= as a key even is received. That would also solve the sticky tooltips =B4featur= e=B4. Best, ------------------------------------------------------------------------ Eric GAUDET Le 01-Mar-2002 a 11:32:55 "Parler pour ne rien dire et ne rien dire pour parler sont les deux principes majeurs et rigoureux de tous ceux qui feraient mieux de la fermer avant de l'ouvrir." ------------------------------------------------------------------------ [Dillo-dev]bug track review From: Sebastian Geerken - 2002-03-01 18:21 Hi, I just took some time to take a look on the bug tracking engine. Here are some comments on a few. > 202 > Impact: soft error > At: run time > Type: render bug > Comments: When an image is loading over a slow connection, and you > start scrolling up and down while it draws, some parts of it may not > get updated. HTReproduce: Any big image will do. For instance, try: > http://www.photo.net/photo/pcd1765/walk-away-44.tcl > WorkedBy: nobody This is a problem of Dw_widget_queue_draw_area, which would quite hard to solve. If somebody is *really* interested in solving this, just ask me for a more detailed explanation. > 231 > Impact: soft error > At: run time > Type: misbehaviour > Comments: dillo doesnt render http://www.fuckedcompany.com the leftmost 5% of > the page appears to be cut off HTReproduce: view http://www.fuckedcompany.com > WorkedBy: nobody Dillo does render it, but below the main part of the page. Same as the following: > 233 > Impact: soft error > At: run time > Type: render bug > Comments: dillo doesn't render side-by-side tables (with align=left) > correctly. Dillo renders one table per line, no matter what. It seems > to be a standard to render such tables on the same line. > HTReproduce: http://www.rti-zone.org/dillo/Html.testsuite/table-side.html > http://www.dpreview.com/ > WorkedBy: nobody Such a behavior would be actually not quite standard compliant, see http://www.w3.org/TR/1999/REC-html401-19991224/struct/tables.html#edef-TABLE: align = left|center|right [CI] Deprecated. This attribute specifies the position of the table with respect to the document. Permitted values: * left: The table is to the left of the document. * center: The table is to the center of the document. * right: The table is to the right of the document. I'm not sure whether we should follow the standard, or common use. Anyway, floats will be used for some other purposes. > 289 > Impact: soft error > At: run time > Type: render bug > Comments: when an image tooltip is display and you scroll the page > with the keyboard, dillo fails to render the new scrolled content. > HTReproduce: Go to dillo home page, move the mouse cursor on the top > image, when the tooltip is displayed, press PgDown: the page scrolls > but the bottom is empty. > WorkedBy: nobody This is a Gtk+ bug, it can be simply be reproduced (combination of tooltips and a GtkLayout, in which is painted). Sebastian