December 2012 - Dillo-dev - lists.mailman3.com

Patch for xembed (text corrected)
by podarcis＠lavabit.com Jan. 14, 2013

Jan. 14, 2013

Hi all, Here is a patch to make embedding dillo into an X-window work. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ diff -r 9bad41ea1dde src/xembed.cc --- a/src/xembed.cc Wed Dec 05 13:03:04 2012 +0100 +++ b/src/xembed.cc Sun Dec 09 10:12:36 2012 +0100 // TODO: Implement more XEMBED support; -void Xembed::show() { +void Xembed::create() { createInternal(xid); setXembedInfo(1); Fl::event_dispatch(event_handler); diff -r 9bad41ea1dde src/xembed.hh --- a/src/xembed.hh Wed Dec 05 13:03:04 2012 +0100 +++ b/src/xembed.hh Sun Dec 09 10:12:36 2012 +0100 @@ -14,10 +14,11 @@ public: Xembed(uint32_t xid, int _w, int _h) : Fl_Window(_w, _h) { - this->xid = xid; + this->xid = xid; create(); }; - void show(); + void create(); int handle(int event); + void resize(int x, int y, int w, int h); }; #endif ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ tested with this FLTK program: ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ #include <FL/Fl.H> #include <FL/Fl_Window.H> #include <FL/x.H> #include <cstdio> #include <iostream> #include <sstream> #include <string> using namespace std; int main( int argc_, char *argv_[] ) { Fl_Window dw( 800, 600, "dillo embedded" ); dw.resizable( &dw ); dw.show(); cout << "dw.xid = " << (unsigned long)fl_xid(&dw) << endl; ostringstream cmd; cmd << "dillo -f -g " << dw.w() << "x" << dw.h() << " --xid " << (unsigned long)fl_xid(&dw); if ( argc_ > 1 ) cmd << " " << argv_[1]; cmd << " &"; cout << "system '" << cmd.str() << "'" << endl; system( cmd.str().c_str() ); return Fl::run(); } ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ What does NOT work is automatic resizing of the embedded dillo window, when the size is changed. This would be nice to have. I do not know enough of the embedding stuff, but I wonder if the static int event_handler() ever gets called with type ClientMessage - at least I could not make it get there.. podarcis

4 11

ui colors again
by corvid＠lavabit.com Jan. 13, 2013

Jan. 13, 2013

Today I dug up the old ui colors patch from last September, ripped out the part that used system colors as defaults that was a problem*, and played around with it a little. I really like configurable colors, so, since I haven't heard anyone say anything like feature freeze yet, I'll make another push for this thing. One difference from the previous code is that I'm using inactive() on the images for inactive toolbar buttons because our gray pixmaps tend not to look so good with different UI colors. It might be good to pull it somewhat closer to the background color, but I'm not sure yet. * Under X, FLTK tries to "Read colors that KDE writes to the xrdb database", which may or may not work fine for those people running KDE, but...

3 20

wizard, switching between UIs
by corvid＠lavabit.com Jan. 2, 2013

Jan. 2, 2013

Yesterday I was playing with ctrl-tab, of course, and I noticed how the state of the focused tab would sometimes change. I went through a few rounds of fixing this in different ways before finding something that didn't seem icky. Detailed comment in patch.

2 3

Render bug
by Hole.destructor＠gmx.de Jan. 2, 2013

Jan. 2, 2013

Dear Dillo developers, sorry for not creating a proper bug report, but the "How to reproduce" message field is too small for the example file. :) The issue ========= After closing the "Save As" dialog, the background color of the HTML page is not redrawn correctly (leaving white rectangles instead of the background color). How to reproduce ================ 1. open the appended html file (note the grey background) 2. click on "file" link 3. move the "Save As" window over the grey background 4. close the "Save As" window --> parts of the background are drawn white, instead of grey Best, Alex

3 3

CAPI ignores download requests when not in cache or handled by the downloads DPI.
by onepoint＠starurchin.org Jan. 2, 2013

Jan. 2, 2013

I wondered why "Save link as" silently failed on some of my local web pages if I had not already visited the link. It turns out they were "file:///" links and "Save link as" sends a download request which the CAPI silently ignores if the URL is not cached and is not handled by the downloads DPI. I added a MSG_WARN to log this but haven't changed any behaviour. I still found it surprising and was wondering if it could be improved? Regards, Jeremy Henty

4 5

hyphenator valgrind error
by corvid＠lavabit.com Jan. 2, 2013

Jan. 2, 2013

==30983== Invalid read of size 1 ==30983== at 0x81151B6: fl_utf8back (fl_utf.c:259) ==30983== by 0x80A2320: dw::fltk::FltkPlatform::prevGlyph(char const*, int) (fltkplatform.cc:599) ==30983== by 0x809EDA3: dw::Hyphenator::hyphenateSingleWord(dw::core::Platform*, char*, int, lout::misc::SimpleVector<int>*) (hyphenator.cc:327) ==30983== by 0x809E9F2: dw::Hyphenator::hyphenateWord(dw::core::Platform*, char const*, int*) (hyphenator.cc:256) ==30983== by 0x809C087: dw::Textblock::hyphenateWord(int) (textblock_linebreaking.cc:762) ==30983== by 0x809BB6E: dw::Textblock::wordWrap(int, bool) (textblock_linebreaking.cc:609) ==30983== by 0x809B6A9: dw::Textblock::processWord(int) (textblock_linebreaking.cc:441) ==30983== by 0x8098753: dw::Textblock::addText0(char const*, unsigned int, short, dw::core::style::Style*, dw::core::Requisition*) (textblock.cc:1694) ==30983== by 0x8097B02: dw::Textblock::addText(char const*, unsigned int, dw::core::style::Style*) (textblock.cc:1478) ==30983== by 0x8073F5C: Html_process_word(DilloHtml*, char const*, int) (html.cc:1216) ==30983== by 0x807A348: Html_write_raw(DilloHtml*, char*, int, int) (html.cc:3923) ==30983== by 0x80727A9: DilloHtml::write(char*, int, int) (html.cc:531) ==30983== Address 0x661e6bf is 1 bytes before a block of size 16 alloc'd ==30983== at 0x402AC41: malloc (vg_replace_malloc.c:270) ==30983== by 0x80A223F: dw::fltk::FltkPlatform::textToLower(char const*, int) (fltkplatform.cc:584) ==30983== by 0x809E86A: dw::Hyphenator::hyphenateWord(dw::core::Platform*, char const*, int*) (hyphenator.cc:223) ==30983== by 0x809C087: dw::Textblock::hyphenateWord(int) (textblock_linebreaking.cc:762) ==30983== by 0x809BB6E: dw::Textblock::wordWrap(int, bool) (textblock_linebreaking.cc:609) ==30983== by 0x809B6A9: dw::Textblock::processWord(int) (textblock_linebreaking.cc:441) ==30983== by 0x8098753: dw::Textblock::addText0(char const*, unsigned int, short, dw::core::style::Style*, dw::core::Requisition*) (textblock.cc:1694) ==30983== by 0x8097B02: dw::Textblock::addText(char const*, unsigned int, dw::core::style::Style*) (textblock.cc:1478) ==30983== by 0x8073F5C: Html_process_word(DilloHtml*, char const*, int) (html.cc:1216) ==30983== by 0x807A348: Html_write_raw(DilloHtml*, char*, int, int) (html.cc:3923) ==30983== by 0x80727A9: DilloHtml::write(char*, int, int) (html.cc:531) ==30983== by 0x8079D85: Html_callback(int, _CacheClient*) (html.cc:3799)

2 2

Dpi_connect_socket retry
by p37sitdu＠lavabit.com Jan. 1, 2013

Jan. 1, 2013

Dpi_connect_socket function from src/IO/dpi.c (also duplicated in test/cookies.c) has retry argument. It is set to TRUE on first call, and Dpi_connect_socket calls itself recursively with retry == FALSE in case of error. This retry function is not working. If connection fails, at line 645 sock_fd is set, but ret is still set to -1. Even if the second attempt to connect is successful, -1 is returned and conected sock_fd is lost. It is possible to fix it by adding "ret = sock_fd", but I think it is better to just remove retries completely. It is unlikely that connecting using loopback interface will fail on first try and succeed on retry. Also, comment says "the server closes sock_fd on error". However, sock_fd should still be closed by client to free allocated file descriptor. OS don't know if Dillo is going to read from it (in which case error such as ECONNRESET should happen) so file descriptor can't be reused. File descriptor should be closed in case of error in a_Dpip_build_cmd too. Patch to fix this function and its calls in src/IO/dpi.c and test/cookies.c is attached.

2 6

show_url pref
by corvid＠lavabit.com Dec. 30, 2012

Dec. 30, 2012

Since it's been broken throughout the dillo3 era, the other day I made it so that you can't disable it -- and now I'm wondering what the value is in even having the pref. Removing Location doesn't give us room for anything more important. It's kind of like having a show_viewport preference.

2 1

PATCH: uicmd.cc prefs.save_dir cleanup
by onepoint＠starurchin.org Dec. 30, 2012

Dec. 30, 2012

uicmd.cc exports a get()/set() interface to the default save directory that is never used elsewhere, and used only trivially internally. It also rereads prefs.save_dir at every save, which is redundant. It would be simpler to have an init() interface that is called once. There would be no change to the user. Patch attached. Thoughts? Regards, Jeremy Henty

2 2

findbar behaviour
by corvid＠lavabit.com Dec. 28, 2012

Dec. 28, 2012

When I want to change the case-sensitivity of a search, I don't like having to switch to the mouse and aim it toward the little checkbox. So I looked up why we disable tabbing to the other widgets (and found nothing). Then a little experimentation showed that triggering shortcuts causes widgets to get focus, which we don't want. With that answered, I tried letting the input do its own callbacks upon Enter or Shift-Enter, and letting everything accept focus. This isn't heavily tested, but I wanted to see how the idea goes over first...

2 8

ctrl-PageUp/Down
by corvid＠lavabit.com Dec. 28, 2012

Dec. 28, 2012

I notice that ctrl-PageUp/Down for cycling through tabs doesn't work when a Fl_Input has focus. I can fix that if anyone cares...

2 2

Ctrl-Tab and tabs
by Johannes.Hofmann＠gmx.de Dec. 28, 2012

Dec. 28, 2012

Hi, it seems that Ctrl-Tab no longer cycles tabs if an input field is focused. I did like the previous behaviour better, where Ctrl-Tab cycled tabs independent of where the focus was. Cheers, Johannes

4 4

penalty_em_dash_*
by jcid＠dillo.org Dec. 28, 2012

Dec. 28, 2012

Hi Sebastian, While checking the penalty_em_dash_* patch, I noticed it would be good to add and comment the hyphenation preferences in dillorc. -- Cheers Jorge.-

2 1

textarea key bindings
by corvid＠lavabit.com Dec. 27, 2012

Dec. 27, 2012

In bug 1112, Alexander asked for ctrl-backspace to work a word at a time. I put together the attached and he reports that it's working well.

1 0

show_tooltip=NO
by corvid＠lavabit.com Dec. 26, 2012

Dec. 26, 2012

tipwin still shows tooltips when show_tooltip=NO.

2 1

vsource/dpip endless loop
by corvid＠lavabit.com Dec. 26, 2012

Dec. 26, 2012

I have a recentish version of the html5 working draft as one page, and it's over 6 megs. If I view source on it, this is rather slow and gives me time to hit Escape and Ctrl-W. Then top shows vsource using all of the processor until I kill it. (gdb) bt #0 0x0804a843 in a_Dpip_dsh_read_token (dsh=0x9e7ae18, blocking=1) at dpip.c:455 #1 0x080491f9 in send_html_text (sh=0x9e7ae18, url=0x9e7ced6 "file:/home/corvid/html5.html.gz", data_size=6402268) at vsource.c:121 #2 0x08049460 in main () at vsource.c:219

2 2

How to link to a specific bug report?
by abe＠deuxchevaux.org Dec. 24, 2012

Dec. 24, 2012

Hi, I just submitted a (forwarded) bug report (#1113) and would like to link to it directly. Is there any query string parameter where I can enter the bug number? Neither "bug_number" like for the volunteer and remove CGI scripts does work nor does what=1113. In case there isn't such a featre yet, I'd be happy if you could at least e.g. <a name="1113">...</a> jump mark. Thta should be quite easy with regards to the CGI script's current output. TIA. Kind regards, Axel -- /~\ Plain Text Ribbon Campaign | Axel Beckert \ / Say No to HTML in E-Mail and News | abe at deuxchevaux.org (Mail) X See http://www.asciiribbon.org/ | abe at noone.org (Mail+Jabber) / \ I love long mails: http://email.is-not-s.ms/ | http://noone.org/abe/ (Web)

2 2

Extensions for hyphenation
by sgeerken＠dillo.org Dec. 23, 2012

Dec. 23, 2012

Hi! At http://flpsed.org/hgweb/dillo_hyphen, you'll find some extensions for hyphenation I've not yet merged into the main repository. Still needs some documentation, but here is an overview: There are now configuration variables for dillorc (see source): penalties for hyphens, as well as the left and right side of an em-dash. The suffix "_2" means that this value is used for lines following a line which ends already with a hyphen. When this value is larger, two adjacent lines ending with a hyphen are avoided. For values, see the definition of the "badness". Typical values: 0 = Penalty used for normal spaces. 1 = A justified line with spaces having 150% or 67% of the ideal space width has this as badness. 8 = A justified line with spaces twice as wide as ideally has this as badness. "inf" may be used (preventing a break in any case); also "-inf" (forcing a break), although the latter makes no sense and may lead to strange results. There is a text page, test/hyphens-etc.html, to play around. Sebastian

4 21

[patch] dns.c: missing break in a_Dns_dillohost_to_string
by p37sitdu＠lavabit.com Dec. 20, 2012

Dec. 20, 2012

Fix is attached.

2 1

config.h.in and configure.in
by onepoint＠starurchin.org Dec. 20, 2012

Dec. 20, 2012

I thought that we agreed[1] to remove config.h.in from the repository and to rename configure.in to configure.ac . Can I make these changes or is more discussion needed? Regards, Jeremy Henty [1]: http://lists.auriga.wearlab.de/pipermail/dillo-dev/2012-December/009618.html

2 2

[patch] send_dpip_tag in vsource.c is unused
by p37sitdu＠lavabit.com Dec. 20, 2012

Dec. 20, 2012

Attached patch removes unused function send_dpip_tag. Functions send_plain_text, send_numbered_text and send_html_text are made static.

1 0

dpid patches
by p37sitdu＠lavabit.com Dec. 18, 2012

Dec. 18, 2012

First patch is for removing expired TODO item, just resending it. Second patch is new, for dpid.c. Function fill_services_list has a bug: on line 463 there is a check *services_list != NULL. If it is true, then -1 is returned, but dpidrc_stream is not closed. Error message on line 456 says "popen failed", should be "fopen failed". fill_services_list checks for availability of user_dpidir and sys_dpidir, but after line 449 these variables are not used. They are just dFreed in the end. Looks like code was just copypasted from register_all. fill_services_list is only called from main.c:248 and dpid.c:807 of dpid.c. register_all is called from main.c:235 and dpid.c:806. So fill_services_list is always called after register_all and doing the same check is useless. On line 503 there is a conversion of services_alpha_comp to type (dCompareFunc). I changed the function services_alpha_comp to the usual style of writing comparison functions so no conversion is required. In the current state it is undefined behaviour. [1] [1] http://stackoverflow.com/questions/559581/casting-a-function-pointer-to-ano…

1 0

[patch] Gopher DPI
by p37sitdu＠lavabit.com Dec. 17, 2012

Dec. 17, 2012

I implemented basic Gopher support using DPI, patch is attached. It is already quite usable. "URL:" links are not supported, but Gopher servers send redirect pages for compatibility. The only major problem is unrecognized file types. I looked at FTP DPI. It sends Content-type: application/octet-stream. This forces Dillo to ask downloads DPI to start download. With gopher it doesn't work as downloads DPI doesn't suppport gopher. So Dillo just says "Content-Type 'application/octet-stream' not viewable.". I left it as it is for now, because I don't think adding support for all known and unknown protocols into downloads.cc is the right solution. Maybe downloads DPI protocol should be changed so Dillo would feed plugin with data it receives as octet-stream and downloads DPI would only save what it receives to disk instead of connecting to server and downloading the file itself.

4 9

Tooltip bugs revisited
by jcid＠dillo.org Dec. 17, 2012

Dec. 17, 2012

Hi there, These days I've devoted some more effort to the tooltips bug [1] and its workaround patch [2]. The motivation is that the current workaround involves a full UI and viewport expose (i.e. a full erase/redraw) that produces annoying flickering when the mouse is in the UI but outside the viewport (e.g. over the panel). Over time I've learnt to move the mouse pointer outside the panel and into the viewport to scroll without flickering (but probably I'm among the few that know this trick). So far (and despite the warning inside FLTK), I've experimented with a custom tooltip handling class, and a couple custom UI widgets, that seem to work well together (i.e. solve the problem without the workaround :) ). Testing revealed the workaround concealed another problem, which I don't remember whether it was detected before, and this is that viewport mouse events get "through" the panel UI!. How?, placing the mouse pointer over the panel (e.g. to the right of the Tools button, or even inside the Location), and scrolling a page full of links under it, will make the mouse pointer change to a hand, and the destination link appear in the status bar area. This bug is revealed by disabling the workaround: just apply the small attached patch. Clearly there's a viewport offset problem somewhere. Correcting it may save us lots of headaches in the future. Does anybody know or has a clue of where this may be happening? TIA. [1] http://lists.auriga.wearlab.de/pipermail/dillo-dev/2011-July/008494.html [2] http://lists.auriga.wearlab.de/pipermail/dillo-dev/2011-July/008515.html -- Cheers Jorge.-

2 6

Fwd: Re: Dillo early exit
by sgeerken＠dillo.org Dec. 15, 2012

Dec. 15, 2012

Somehow this post got lost ...

3 3