- Dillo-dev - lists.mailman3.com

hairy bug in IO fixed!
by Jorge Arellano Cid Oct. 31, 2003

Oct. 31, 2003

Hi there! The last five or six days were spent on trying to find (& fix) a bug that caused the CPU to hog under very hard to reproduce conditions. I know that heavy duty dillo users had noticed that in rare ocassions dillo starts taking all the CPU, and that's almost impossible to reproduce it again with another dillo... Well, the first interesting thing is that the CPU hog is "soft" (i.e. doesn't block dillo) because it is caused by a polling loop. The other surprising fact is that the CPU-hog occurs outside of dillo (!). That is, not in any part of its code. So if you were to attach GDB to find the bug (once on hog), you'd be always pointed outside dillo's source, but with a clue: polling. I was working with the ftp dpi when I finally came accross a way to reproduce the bug, so I took the chance and started digging. A whole revision/tweaking of the IO engine close/abort handling, a revision of glib's sources (gio channels) to finally find a way to stop the polling loop. Short story, there were several involved parts: threads, the IO, external processes, several open connections in parallel, glib (and the kernel). It turned out to be a race condition in the window between the actual close of the file descriptor, the glib source event removal for the watch, and the kernel reuse of FD number. The source code for gio channels (inside glib) hinted a way to remove the event source cleanly, but outside the gio channel's explicit API. The details are in IO.[ch] in CVS! Cheers Jorge.- PS: Indan wrote: > 3) There is no efficient way of (un)registering file descriptor events in > GTK. (FLTK has a perfect interface). Yes there is! (subtle but effective) > Did anyone found the bug yet, without reading the diff? If not, then it's > as harmless as I thought ;-). Still not nice of course, but it's unclear > where it goes wrong exactly, doesn't seem to be easily fixed without > changing Dillo a lot. I hope this patch and explanations helps you with point number 3. More comments later... Cheers 2 Jorge.-

1 0

Re: [Dillo-dev] HTTPS Patch
by David McKee Oct. 28, 2003

Oct. 28, 2003

Folks, When I apply and attempt to compile (on a Redhat 6.2 machine, gcc 2.95), I have two problems. First, the build doesn't know where to find my openssl headers and libraries, and there does not seem to be a ./configure option for this. (BTW: My goodies are in /usr/local/ssl/...) I can fix this locally, but a config option might be in order. Secondly, I get two compilation errors in IO.c. Both are related to the code: errno = 0; ssize_t St = SSL_read(io->ssl, io->Buf, io->BufSize); which I believe should (strictly) be ssize_t St; errno = 0; St = SSL_read(io->ssl, io->Buf, io->BufSize); because declaration should come before statements. In anycase, making this change in both places (around lines 822 and 852) fixes the problem for me. Now to try it out... Thanks, -- -- David McKee -- dmckee(a)jlab.org -- (757) 269-7492 (Office)

5 6

HTTPS patch
by Indan Zupancic Oct. 27, 2003

Oct. 27, 2003

Get it here: http://www.xs4all.nl/~dorinek/dillo/dillo-ssl.diff.gz It's not as nice and simple as I wanted it to be, thanks to three things: 1) SSL can always read and write, meaning that SSL_write may want to read. 2) Dillo uses two IOData structures per file descriptor. 3) There is no efficient way of (un)registering file descriptor events in GTK. (FLTK has a perfect interface). This three facts contribute a lot to the complexity of the code. There is one known bug, you may try to find it. ;) Dillo will not crash, I made it exit with a message. This patch adds also generic ssl/tls support to Dillo, most work was done in IO.c, http.c has hardly changed. The current way to tell the IO engine that it should use ssl is by giving a negative fd to a_IO_new. This avoids extra interface, is simple, and doesn't clash with anything else. Dillo is about 2K bigger after the patch. It was made against the current cvs, apply it with 'patch -p1 < dillo-ssl.diff' while you're in the dillo dir. Run configure with --disable-ssl if you don't want https... Have fun. p.s. I hate glib ;), it's totally worthless, the most complex piece Dillo uses is a hash table, and glib implements it as a linked list, look in the code yourself if you don't believe me. Since when is ansi c not good enough? Please switch to FLTK. :) p.s.s This mail was sent from webmail with a patched Dillo. :)

2 3

fltk c binding
by cesar Oct. 27, 2003

Oct. 27, 2003

Fltk was modeled following XForms and has a XForms compatibility layer: http://fltk.org/documentation.php/doc-1.1/forms.html#forms http://fltk.org/documentation.php/doc-2.0/forms.html#forms You can learn more about XForms here: http://world.std.com/~xforms/ pd: I don't know how to program using Fltk or XForms. I only was aware of this compatibility layer. Both also have a GUI designer. I only know of very few apps using XForms, the most important of those being LyX. cesar;

2 1

[PATCH] Version 2 of keyboard navigation, solves problem with Tab
by Frank de Lange Oct. 26, 2003

Oct. 26, 2003

Hi'all, The next version (2) of my keyboard navigation patch is ready. See the end of this message for download instructions. Do note that the tab/frame patch already contains this functionality, so there is no need (and no way) to apply this patch on top of that or the other way around. This release does not contain any major new functionality. It solves the problems with Tab, which now does what you'd expect it to do: cycle through form widgets. For more info on what this patch does, see the last message on the list or the doc/Keyboard_Navigation.txt document (after applying the patch, naturally...) Download instructions ===================== There is currently only one version of the patch, made against 20031017/11:26GMT CVS. As always, have a look at the patch website (http://www.geocities.com/ikbenfrank/) for the latest version of the patch. This patch does NOT apply on top of the tab/frame patch, and the tab/frame patch does not apply on top of this one! The tab/frame patch already contains this functionality, so it is not needed. http://www.geocities.com/ikbenfrank/dillo-200310171559-kbnav.patch.gz size: 14494 bytes (gzipped, as downloaded) 64277 bytes (uncompressed) md5sum: 8ebd2450b762a720363346c0c4a7a537 (gzipped, as downloaded) 9538b60172b910e89c993d887b28bc97 (uncompressed) 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." ]

8 24

Bug #502 solved
by Frédéric Bothamy Oct. 24, 2003

Oct. 24, 2003

Hello, I submitted the bug #502 a while ago about some PNG images not displayed correctly. And today someone finally gave the explanation: the Debian web server sends GIFs images with image/png content type (and png extension) (for more details, see http://lists.debian.org/debian-www/2003/debian-www-200310/msg00133.html) Could someone, please, close this (non-)bug? Sorry for not testing this issue enough and thanks for this wonderful, lightweight browser. Fred PS: for anyone interested in using Dillo 0.7.3 with tab+frame patch (v10) under Debian woody (i386 only), I have an unofficial package available at http://olympie.dyndns.org/debian/debs/dillo_0.7.3-1.3_i386.deb.

2 1

frames, tabs patch
by Jens Arm Oct. 24, 2003

Oct. 24, 2003

Hi When will the frames- and tabs-patch integrated into CVS ? Jens

1 0

Re: [Dillo-dev] FLTK
by Jorge Arellano Cid Oct. 23, 2003

Oct. 23, 2003

Hi there! Just to add more information about FLTK: http://mail.gnu.org/archive/html/texmacs-dev/2003-02/msg00073.html Very informative. Cheers Jorge.-

1 0

Can't find dpidrc
by Nigel Wilkinson Oct. 23, 2003

Oct. 23, 2003

Hi folks I've just built dillo-0.8.0-pre from the CVS but when I try to use bookmarks it can't fimd dpidrc either in /usr/lib/dillo/dpi or in the ~/.dillo/dpi directory (and its right, they aren't there. Where is it or what is the format so that I can create it. Cheers Nigel

3 5

Re: [Dillo-dev] FLTK
by Jorge Arellano Cid Oct. 23, 2003

Oct. 23, 2003

Hi there! After the previous answers from Michael Sweet (forwarded to this list), Sebastian rose a key question: > GtkLayout is mainly there to overcome the limited size of X > windows. When I started to develop Dw, I first tried to get a similar > result by writing anything myself, but finally gave it up. Look at the > Gtk+ documentation, GtkLayout does some weird things which may not be > possible with FLTK. Once again, here're the answers. Cheers Jorge.- PS note: Fl_Help_View is a widget for basic HTML rendering. ---------- Forwarded message ---------- Date: Wed, 22 Oct 2003 16:03:52 -0400 From: Michael Sweet <mike(a)easysw.com> To: Jorge Arellano Cid <jcid(a)dillo.org> Subject: Re: FLTK and Dillo Jorge Arellano Cid wrote: > ... > One interesting concern raised by Sebastian is about a way to > overcome the limited-window-size problem of X windows (GTK+ uses > GtkLayout to overcome it). I know FLTK runs on WIN32 so it should > have no problems with large windows, but I couldn't find any > formal doc stating it. This is an issue for us because of large > scrolled html pages (or should I say books?). Are you referring to the 16-bit coordinate issues? If so, those are also present on WIN32... FLTK 1.1.x uses 16-bit integers for coordinates, so to support views larger than 16-bits you'll need to do a little work yourself. For example, Fl_Help_View offsets the X,Y values by the current scroll position and does gross clipping of blocks to support really long HTML files. FLTK 2.x uses 32-bit integers for coordinates and doesn't have this limitation. It will also be providing device classes to support X11, PostScript, PDF, etc. rendering in a device- independent way. >>> 2.- Unicode, UTF-8. >>> This was the main reason for considering GTK2. Is FLTK ready >>> to render Unicode characters? >> >>The patched version is, and FLTK 2.0 will fully embrace UTF-8. > > > Do you mean current FLTK 2.0 CVS does not have the patch yet? AFAIK, 2.0 doesn't have it yet. Bill Spitzak is probably a better person to talk to on that issue, as I've been sticking to maintaining 1.1.x (lack of time on my part - too many other projects! :) > If so, can it be patched or should we use 1.1.4 + utf-8 patch? > > In other words, which version should we start testing with? > What's your advice to structure it (i.e on which widgets should > we base HTML rendering)? Use 1.1.4 + UTF-8 patch; 2.0 is still in a state of flux, and until we have the new device classes in place there may still be some API changes. If you write to 1.1.x now, the 1.x compatibility headers in 2.0 should be enough to support your apps with either release. You'll probably want to subclass Fl_Group and use the existing Fl_Scrollbar classes to provide your scrollbars. The Fl_Help_View widget might serve as a guide for your porting effort, as it supports a limited subset of HTML and does all of the scrolling, linking, etc. stuff. -- ______________________________________________________________________ Michael Sweet, Easy Software Products mike at easysw dot com Printing Software for UNIX http://www.easysw.com

1 0