- Dillo-dev - lists.mailman3.com

Dillo Win32 Thoughts
by obeythepenguin＠gmail.com Oct. 18, 2011

Oct. 18, 2011

By the way, if you want to try Dillo-Win32 on Linux/Unix, download and apply this patch before you compile it: http://dillo-win32.sourceforge.net/TODO/dfcntl.diff This fixes the dFcntl implementation on Unix, so nonblocking I/O works correctly and the browser won't lock up. I only noticed this bug today, but I'm going to put up an updated release with the fix included when I get a chance. Apologies for any inconvenience, ~Benjamin

1 0

Dillo Win32 Thoughts
by obeythepenguin＠gmail.com Oct. 18, 2011

Oct. 18, 2011

Georg wrote: > Benjamin, > you mention that your Win32 version will also compile with Linux. I > tried to compile it on Linux and did not have much success. Your source > code package does not include a configure file for that such as the main > version does. I have doubts too that you frequently test if your version > still works on Linux. You can generate the configure script yourself by running autogen.sh (requires autoconf and automake). I don't include a pregenerated configure script mainly because I build my source tarballs using 'hg archive', and Mercurial is set not to track those files. I may change my release process in future versions to make installation from source easier. Technically, you're correct that I don't test Dillo-Win32 on *Linux*. I do, however, use it quite often on OpenBSD, as you can see in this (quite large) screenshot from my development machine: http://obeythepenguin.users.sourceforge.net/openbsd/files/screenshot.png > Will this version when compiled for Linux also use the sockets > abstraction layer and not use the dpi parts? > Georg The sockets abstraction layer and DPI are two entirely separate things. The sockets layer is just some low-level network code that lets Dillo connect whether it's on Unix or Windows. Dillo doesn't know the difference, and that's the entire point -- it behaves exactly the same on both systems. As far as DPI, it's a little more complicated. In a nutshell: - Bookmarks, cookies, and HTTPS are built into the browser, and don't use DPI. - Downloads are built in by default, but you can disable the built-in downloader and use the DPI. - Everything else will fall back on DPI if it's available. Dillo-Win32 doesn't build dpid and DPI programs by default, but it will use mainline's DPIs if they're installed on the system. I haven't removed DPI support from Dillo -- I just moved most of their functionality into the browser. Hope this helps, ~Benjamin

1 0

Fwd: RE: keysrc broken in dillo 3?
by obeythepenguin＠gmail.com Oct. 18, 2011

Oct. 18, 2011

Here's an interesting one for you guys... I had a user email me this morning saying his keysrc file stopped working after he upgraded from Dillo 2.2 to 3.0. Neither his existing keysrc nor a fresh one would work on his machine, while I couldn't reproduce the problem on mine. We just now discovered by chance it wasn't his keysrc that was the problem, but -- strangely enough -- that he had set start_page=about:blank in his dillorc. I've forwarded his original message below (with names and headers removed for privacy), which includes a more detailed explanation. I was just curious if anyone could reproduce this in mainline before I filed a bug report, or if this behavior was specific to Dillo-Win32. I don't have a Unix system handy now, or else I'd just test it myself. :-) Cheers, ~Benjamin ------- Forwarded message ------- I tracked down the source of the behavior and I think it should be reproducible now. In dillorc I had the following line: start_page=about:blank (By now you've probably figured out I'm a Firefox user..) Apparently Dillo 3 does not like for start_page to have this value, and will only load keysrc after a non-blank page is loaded?! This doesn't happen in Dillo 2.2 (20110630) but I'm not sure if this is an intentional change in behavior; probably not, considering the following weirdness: These values allow keysrc to be loaded without having to load a non-blank page: start_page=about:blank#any comment start_page= Sorry, I haven't had a chance to try any of this out in Linux, so still not sure whether this is an upstream regression.. For completeness my old cookiesrc, dillorc, and keysrc files are all attached (files that reproduce the artifact).

1 0

[patch]: download window resizes its contents to its width
by onepoint＠starurchin.org Oct. 17, 2011

Oct. 17, 2011

Purely cosmetic, but I dislike it when the contents ignore their container, even in only one out of two dimensions. Comments? Regards, Jeremy Henty

2 4

Scrollable view without scrollbar
by Hole.destructor＠gmx.de Oct. 17, 2011

Oct. 17, 2011

Dear Dillo developers, I found a site today (see appendix), which is scrollable in Dillo, i.e. you can press <left> and <right> and the view scrolls left and right, but the view has no horizontal scrollbar. What is also strange here is that the site has no fixed width elements, so the view should not be scrollable. What's happening here? Kind regards, Alexander P.S.: The site is a boiled down version from today's blog.fefe.de

2 2

small panel_size, small_icons, progress boxes don't fit.
by corvid＠lavabit.com Oct. 17, 2011

Oct. 17, 2011

Just noticed that with small panel_size and small_icons, the labels in the progress boxes don't fit.

1 0

Dillo Win32 Thoughts
by mailbox＠georgpotthast.de Oct. 17, 2011

Oct. 17, 2011

Benjamin, you mention that your Win32 version will also compile with Linux. I tried to compile it on Linux and did not have much success. Your source code package does not include a configure file for that such as the main version does. I have doubts too that you frequently test if your version still works on Linux. Will this version when compiled for Linux also use the sockets abstraction layer and not use the dpi parts? Georg

1 0

Dillo Win32 Thoughts
by obeythepenguin＠gmail.com Oct. 17, 2011

Oct. 17, 2011

Hmm, looks like I picked the wrong week to unsubscribe... (it's been a busy semester so I'm trying to cut back on distractions and inbox clutter) Roger wrote: > I had the pleasure of trying out Dillo win32 and I must say I like it. > (Itstarts a heck of alot faster then that bloated green lizard thing ... > sorry,the name doesn't come to mind right now. ;-) >Only one or two things. >1) When I first installed it. I was completely stumped as to how to use > thebookmarks. I was trying to add a bookmark without going to the > actual sitefirst. One thing that could really eleviate all this, is by > setting a fewpreset default bookmarks (ie. www.google.com, and other > favorite searchengines.) Notice, there's no "add bookmark" within the > menu and you have tohave the insight to right click on the displayed > page to get this menu.) Thanks for the suggestion -- I'll add some presets in the next release. One small correction: there *is* an "add bookmark" option, but it's labeled "Bookmark this page" for consistency with the right-click menu. You can also press Ctrl-D to get it, like in Firefox or Opera. I based my bookmarks interface on Opera's, since that's what I'm personally most familiar with (I think Firefox's is similar). So it makes sense to me, but it's probably a bit of a change from mainline Dillo's. I'll have to see what other users think so I know if I need to tweak it. > 2) Copy & paste work only within the displayed page and not within the > urlfield. (Although I really miss middle button pasting in Windows.) The keyboard shortcuts (Ctrl-C and Ctrl-V, respectively) should work for both the page and URL field. Windows doesn't support X's select-and-middle-click mechanism. I agree that it would be really nice to have -- that and a decent terminal emulator like aterm would make my life so much nicer. > 3) One thing I really dislike about Windows, there's no main package > managerlike the Linux distributions for updating all of your other win32 > open sourcepackages. Think most packages & drivers have been all > implementing their ownupdate check routine within their Help menus, > etc. Now, they just have to getthe idea of updating their own packages, > such as Adobe Reader. That would be nice too, although I'm not a fan of the Linux distribution system where you're depending on the distribution maintainers rather than the individual package developers to provide updates. Of course, that all goes back to the lack of proper binary compatibility on the platform, which is a completely different story altogether. > Dillo win32, installs within a few seconds or so. Dillo starts within a > fewseconds too. Dillo win32 is functional as well. Glad you like it! Georg wrote: > what I like about the Win32 version is the fact that it compiles to one > standalone executable file. No dpi is used. > > Georg I rather like that, too ;-) Stephan wrote: > Hello, >I just gave Dillo another try on Cygwin/X11 and got it working at > least partially with cygwin 1.7.9. >You need to compile FLTK using Cygwin libraries and X11: > ./configure --enable-cygwin --enable-x11 && make install >Because the dpis are installed with ".exe" extension (eg > hello.dpi.exe) they aren't found by dpid. The extension is added by > automake: http://www.gnu.org/s/hello/manual/automake/EXEEXT.html > The Makefiles look like this: bookmarks_PROGRAMS = bookmarks.dpi$(EXEEXT) >I don't know how to break automake of doing this for the dpis only, so > I modified dpid to use .exe-extension on cygwin (see attached patch). You can't get rid of the ".exe" extension, because Windows relies on the extension to identify executable files (rather than file permissions, like on Unix). > After starting the XServer (/usr/bin/XWin -multiwindow) you can run > Dillo. Plain browsing with http using proxy and the bookmark & cookie > dpis does work. The other dpi currently fail. >- hello-dpi shows the dialog and without pressing any button the > resultpage (Answer received: Window was closed). After pressing Yes > dpid logs "ERROR: [a_Capi_dpi_send_data] No open connection found" > - vsource just crashes > - file, ftp and https can't connect: > Dpi_get_server_port: can't read server port from dpid. > ** ERROR **: dpi.c: can't start dpi daemon >Looks like the socket implementation of cygwin still has some quirks > which can maybe tracked down. First of all, I'll just say from my own experience: sockets on Windows are a mess. The main problem is socket and file handles aren't interchangeable on Windows, so you can't e.g. dup2() stdio handles to a socket. DPI relies heavily on that trick, which is the main technical reason there's no native DPI on Windows. As I understand it, Cygwin implements a bunch of clever tricks to emulate the Unix sockets behavior, but they're quirky at best because of technical differences in the underlying system design. > Is there any interest to support this target on a source code basis? >Thanks and Greets, > Stephan From what I've seen, the Dillo developers aren't interested in Windows at all, via Cygwin or otherwise. Of course, I don't want to put words into their mouths; Cygwin will probably require less extensive patching than a native port like mine, so you may have better luck than I have. I will say I'm personally not a fan of Cygwin, since it negates all the advantages Dillo provides: speed, efficiency, small size, easy installation. A native port like Dillo-Win32 doesn't involve any of Cygwin's complexity or overhead, so it's much more useful to the vast majority of Windows users. Side note: I'm working on a patch to provide native DPI on Windows. So far it compiles, but it's not functional yet, and it's *extremely* ugly code. Email me if you want a copy. Georg wrote: > I suggest to implement a configure option that causes the dpi functions > tobe linked with the main dillo code. The dpic then needs to be adapted > tocall these functions directly instead of forking a new process. This > wouldgreatly simplify porting Dillo to other platforms. >I guess the Dillo developers have already done this for their > development sothey can step through the entire code with GDB. >Georg It's not quite as simple to do as you've described, but that's essentially what I've done with Dillo-Win32: move all the DPI functionality into the browser. Some of it, like HTTPS support and the downloader, is new code, but a lot of it is straight copy-and-paste -- for example, cookies, and the bookmarks backend (but not the user interface). I've also introduced some new features like a sockets abstraction layer that greatly help Dillo's portability. Dillo-Win32's a bad name for my project, because it's not _only_ for Windows, but since that was my original target the name stuck. I'd like to get my code into mainline (that was my goal from the start), but it requires extensive changes to Dillo's inner workings, and the DPI thing has been a bit of an issue. I'm not absolutely opposed to it, but I think the current implementation is deeply flawed, and I don't believe in implementing required functionality through an extension mechanism since required functionality is by definition not an extension. Anyway, sorry for the _obscenely_ long email, but I hope this covers everything. :-) Thanks again for all the feedback, suggestions, discussions, distractions, etc. Cheers, ~Benjamin

2 1

text selection and images
by Johannes.Hofmann＠gmx.de Oct. 16, 2011

Oct. 16, 2011

There seems to be a problem with images and text selection. When you select the text in foo bar <img src="foo" alt="img"> foo bar the second foo gets selected completely while the mouse is in the space after the <img>. The relevant code seems to be Textblock::sendSelectionEvent(), but I couldn't spot the error yet. Any ideas? Cheers, Johannes

2 2

Dillo Win32 Thoughts
by mailbox＠georgpotthast.de Oct. 15, 2011

Oct. 15, 2011

I suggest to implement a configure option that causes the dpi functions to be linked with the main dillo code. The dpic then needs to be adapted to call these functions directly instead of forking a new process. This would greatly simplify porting Dillo to other platforms. I guess the Dillo developers have already done this for their development so they can step through the entire code with GDB. Georg

2 1