- Dillo-dev - lists.mailman3.com

Latest parser changes

by Jorge Arellano Cid

Hi there! After all the trouble with trying to convince Slashdot's webmasters to fix their HTML (because of the loftier importance of adhering to standards), and considering Andreas S. comment on that dillo was making a weird rendering only, but not helping to find the cause to fix the problem, I started a more in-depth review of it. As time passed the code I was developing turned more and more interesting and finally here you have it! This is a quote from the ChangeLog: * Adapted the generic parser to make HTML error detection, providing the line number and a hint (expected tag) in the error message! * Added 'show_html_wanings' option to dillorc (boolean). * Added information about optional, required and forbidden end tags. * Modified the parser's handling of closing tags to account for elements with an optional close tag, and for more accurate diagnosis messages. * Added 'use_old_parser' option to dillorc (boolean). * Fixed the handling of HEAD and BODY elements to account for their double optional condition (both open and close tags are optional). This basically means that now dillo will detect and print useful error messages about bad-formed HTML. After using it for a while it helped me to find bugs I have overseen for months and it proved very accurate and helpful! (it even catched one at the W3C site! :) I hope it will be a good tool for webmasters, and for people authoring simple web pages. In our side, it will help to polish our HTML parsing. For instance, it soon became clear that we're sometimes pulling tags when they should stay. For instance: <p> <ol><li> One <li> two </ol> </p> <- detects an error here. because <ol> anticipatedly cleans <p> from the stack. This may look not relevant, but when trying to build a DOM tree it certainly is! Some comments: The new parsing behaviour makes sites render different from what they used to. Sometimes they show better, some others worse. After all it is bad-formed HTML what we're dealing with. For instance: heise.de a bit worst but usable. slashdot.org renders ok. sf.net worst but usable. Please bear in mind that now dillo can diagnose these errors at render time so, as a user, now you can send a meaningful error report to a site's webmaster! (for instance slashdot is diagnosed on-the-fly). Our experience is that most webmasters welcome this reports. Finally, remember that you can turn off the HTML error messages show_html_wanings=NO and that you can fall back to the old parser: use_old_parser=YES Enjoy! (I hope ;) Jorge.-

Nov. 11, 2003

[RFC] Quo Vadis Dillo

by Frank de Lange

Hi'all, Now that the last release of the patches seem to work (number 13 was less of a success...)... Where do we want to be tomorrow? What needs to be added/changed to improve Dillo? Should the tab/frame/etcetera + Indan's patch + ??? be forked into an 'experimental' version, somewhat like the Sylpheed MUA was forked into Sylpheed-claws? Talking about my patch, what does need improvement? Some ideas: - select contents of location bar/search field when they are focused (DONE, in next version) - make the 'search the web' popup disappear when popup_dialogs=NO, replace it with a non_popup version or use the location bar for search (like several other browsers do) - more key bindings (replacing current key handling where applicable) - remove compile options for tabs and frames, make tabs/frames standard? [ PDA/embedded developers: is this OK with you? ] - ??? Cheers//Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Nov. 11, 2003

Re: Keybd 3 patch

by Frank de Lange

Jorge Arellano Cid wrote: >Hi Frank, > > I just noticed the newes keyboard navigation patch in the >mailing list. Please tell me if this problems were fixed or if a >should wait the next version before aiming to a merge: > > These problems are gone. I have completely overhauled the internals so that focus always is in sync with the rest of Dillo. Which just happened to fix the problem Sebastian noticed... I'd say, wait a week or so to see if any issues crop up. If all seems OK commit the patch. [ LIST: please give keyboard navigation, on either version of the patch, a workout. This needs some testing if it is to be part of mainstream Dillo. Try it in frame sites, try navigatin between frames with Tab/Shift-Tab, try the keyboard bindings, etc. ] By the way, the patch also contains a fix to a problem with dpid/Makefile.am (which did not honour DESTDIR on installation, messing up make install for non-root users). Let me know when you apply this, because it will affect the tab/frame/kbnav patch s well. Cheers//Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Nov. 11, 2003

Re: [Dillo-dev] HTTPS Patch

by Paul Pelzl

On Tue, Nov 11, 2003 at 10:37:41AM +0000, robert w hall wrote: > To reiterate > It would be _really nice_ if all these links to these very useful > patches were findable from the dillo main site :-) Perhaps some enterprising individual (not me) would be interested in hosting some sort of Dillo Wiki. Any patch authors could then post links on the wiki without having to bother website maintainers... Paul

Nov. 11, 2003

Re: [Dillo-dev] HTTPS Patch

by Indan Zupancic

> Note: When Indian continues to maintain his https patch i'll probably do > not bother with mine, lazy as i am... Also the Indian's idea of cleaning > up ccc into something more easy to understand is nice (current ccc has at > least one bad property - it is _hard_ to understand what goes on in it). It is my intention to keep my https up to date, it will be in a couple of days. I'm more or less waiting till Jorge tells me that I may go ahead with writing a patch that does that IOData and GDK stuff, or that he wants to keep it as it is now. That way it's clear for me what I can do, and how the more final version of my https patch should look like. I started documenting how the CCC functions are used currently in Dillo. After having it clear what happens now it's easier to see how to improve it, and what's wrong with the existing implementation. I do not know how hard it is to manage an experimental CVS branch, it has advantages and disadvantages. One pro is that the patches always work with eachother, because they use the same code, but that also means that the patches won't work seperately, that is, there is just an extended Dillo version, and no real patches anymore. It is very good to work with a stable code base, because then it's clear that all the encountered bugs are caused by your code, and not by someone else's code. Also it's useful when the base code doesn't change that much, especially when the patch is young. It is easier to keep the code up to date when it has it's own cvs branch, but on the other hand it could also mean that if someone drops his code and stops working on it, that someone else need to keep it up to date. Altogether quite risky. Of course everything is much simpler when only a few people use the alternative branch, it shouldn't be a dump place for never updated patches (tiny ones are ok of course). I think it will be only succesful when the branch is used only for things that have a chance to come in the official Dillo, as a sort of testing and developing place. Otherwise it will be very hard to keep it in sync with the official Dillo version, and the changes will only become greater after a while. Greetings, Indan

Nov. 11, 2003

Re: [Dillo-dev] HTTPS Patch

by robert w hall

In message <Pine.LNX.4.58.0311111201430.31566(a)kiku.itsise>, Madis Janson <madis(a)ats.cyber.ee> writes >you can use my patch for 0.7.3 until something better appears... > >http://www.zone.ee/myzz/dillo/0.7.3/dillo-0.7.3-mzz-wtab.diff.bz2 >http://www.zone.ee/myzz/dillo/0.7.3/dillo-0.7.3-mzz.html > To reiterate It would be _really nice_ if all these links to these very useful patches were findable from the dillo main site :-) -- robert w hall

Nov. 11, 2003

Re: [Dillo-dev] HTTPS Patch

by Madis Janson

On Tue, 11 Nov 2003, Mike Eriksen wrote: > Sorry to interrupt... I'm a keen user of Dillo but no hacker. I'll not > comment on all the internals of Dillo since you guys are much more clever > on this than me. > > But, see - the reason why I like Dillo is that it is light and mean - and > can be tweaked into something I need. My need is a *light* browser which > can be fitted into a thin client but still allow you to read web mail. > This requires frame and https support. Other users of Dillo, like PDA > users, couldn't care less about frames and https. > you can use my patch for 0.7.3 until something better appears... http://www.zone.ee/myzz/dillo/0.7.3/dillo-0.7.3-mzz-wtab.diff.bz2 http://www.zone.ee/myzz/dillo/0.7.3/dillo-0.7.3-mzz.html It contains both tabs patch v6 and one https implementation (mine), but you should read sslgw/README to get it work, if you want to use it. Note: When Indian continues to maintain his https patch i'll probably do not bother with mine, lazy as i am... Also the Indian's idea of cleaning up ccc into something more easy to understand is nice (current ccc has at least one bad property - it is _hard_ to understand what goes on in it). > The key point here is that neither I nor the PDA guys really have > anywhere else to go than Dillo. >

Nov. 11, 2003

Re: [Dillo-dev] HTTPS Patch

by Frank de Lange

Indan Zupancic wrote: > Having a central place for all Dillo patches is always useful, because > that way it's easier to find and use them for users. Centralised isn't the > same as semi-official, it's a united effort of the patch makers. And that > means they probably have something betters to do, say, improving their > patch and keeping it up to date with cvs. ;) Another thing we can do is go the 'sylpheed-claws' route: create an extended version of Dillo, which regularly syncs up with Dillo CVS (and, possibly, the other way around). That way it is easier to create and maintain those features which are missing from Dillo. The risk is of course that the extended version starts to divert so far from CVS that it is hard to use it as a base to patch CVS. I've thought about going that way with the tab/frame/keyboard_navigation patch, as it does not get that much exposure currently. I have not done it yet, but given the number of patches which languish because of lack of exposure it is probably a good start. An 'experimental' branch on CVS and the main Dillo site would work as well of course... Cheers//Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Nov. 11, 2003

Re: libpng detection again...

by William Trenker

Jorge Arellano Cid wrote: > <candidate line> > CPPFLAGS="$CPPFLAGS `$PNG_CONFIG --cppflags`" > </candidate line> Yes! Hooray! I've been struggling with this for months and at last I have Dillo back again. After applying this change to configure.in and re-running autoconf the ./configure worked and I was able to build Dillo without a hitch. Thank you very much. And thanks Dillo team for a great browser! Regards, Bill

Nov. 11, 2003

Re: [Dillo-dev] HTTPS Patch

by Indan Zupancic

> When I tried to add Indans https patch, it > didn't work. Too old. Two weeks ago I tried the same. At that time the > problem was the opposite! Plain bad luck, Jorge just changed those files enough, and I was bussy with making a pc stable and upgrading it, all my Dillo files are backuped on another pc now, it's a slight mess. :) Also I want to merge the read/write IOData structures, and change to the GDK API, if Jorge and others agree with that. If that's going to change then updating my https patch now is a waste of time, because relatively much is going to change (the https patch will shrink incredibly :). > But in my humble oppinion either Dillo should release official minor > version (0.7.4, 0.7.5 etc) more often so privateers can offer patches > against these, or go along with this suggestion of Indans and support a > semi-official patch model. More often is not the point, Dillo is relatively close to 0.80 and far away from 0.73, so this seems to be just an unfortunate time for patches. Normally you should just use the 0.73 patches. There is no 0.73 https patch from me because I'm not happy with the current implementation, caused by all the hassle with keeping the two IOData structures synced and GTK's io channel annoyance. If there are fewer official Dillo versions it's easier to keep all the patches up to date. Actually, all what is needed is a cvs tag for each stable release, and maybe some extra tag for sort of beta versions that can be used for patches if it turns out that the official releases are not often enough for the patches. This way it's clear for the patches what Dillo version to use, so all can be relatively easily kept synchronized (at least for what Dillo version the patches use). Having a central place for all Dillo patches is always useful, because that way it's easier to find and use them for users. Centralised isn't the same as semi-official, it's a united effort of the patch makers. And that means they probably have something betters to do, say, improving their patch and keeping it up to date with cvs. ;) Greetings, Indan P.S. Seems like I just can't write short mails. :(

Nov. 11, 2003