- Dillo-dev - lists.mailman3.com

[patch] fix redraw issue on resize
by Johannes.Hofmann＠gmx.de May 7, 2008

May 7, 2008

Hi, attached patch fixes an obvious problem with the redraw optimization that results in artifacts when resizing the window. There may be other issues left. Cheers, Johannes

2 1

CVS glitch
by onepoint＠starurchin.org May 7, 2008

May 7, 2008

Just so you know, for some reason it took a long time for the latest patches to show up in CVS. Normally I update more or less as soon as I see Jorge's "committed" message and the patches appear, but this time I saw his acknowledgement late yesterday afternoon and the patches still weren't there by midnight. They turned up this morning so it was not a permanent failure. Did this affect anyone else? I've never seen it before! Jeremy Henty

2 2

[patch]: Refactor {FltkListResource,FltkOptionMenuResource}::sizeRequest()
by onepoint＠starurchin.org May 6, 2008

May 6, 2008

A simple refactoring: FltkListResource and FltkOptionMenuResource use the same code to determine the maximum string width. Jeremy Henty

2 2

[bugfix]: FltkSelectionResource leaks memory
by onepoint＠starurchin.org May 6, 2008

May 6, 2008

corvid pointed this out to me. Patch attached. Jeremy Henty

2 1

Test
by sgeerken＠dillo.org May 6, 2008

May 6, 2008

Hi, this is just a test message. Sebastian

1 0

ListResources: version 3
by onepoint＠starurchin.org May 6, 2008

May 6, 2008

I am attaching two patches that get ListResource selection working properly, ie. you can click items as you wish and FltkListResource::isSelected() will now return the correct value. The first patch is my previous selection patch, unchanged. The second is a small fix that I proposed in the lout::misc::StringBuffer fix thread: it makes FltkListResource::addItem call Widget::set_selected() on fltk::Item widgets, and the call to Menu::set_item() becomes conditional on a virtual boolean hook that defaults to false, but is redefined to true in FltkOptionMenuResource() . The second patch looks bigger than it really is: most of it is just reindenting a block of code that moved inside an "if" statement. (And diff doesn't help by insisting on choosing an unnecessarily complicated set of changes.) It's not a pretty hack, but it's simple and I am wary of making deeper changes until I understand the code better. Please comment, Jeremy Henty

4 5

Q: should we select option groups
by onepoint＠starurchin.org May 6, 2008

May 6, 2008

FLTK allows both Choice and MultiBrowsers to select group items as well as normal leaf items. The Dillo Selection widget currently ignores selected group items, which could be confusing if a user selects a group and expects the application to do something about it. On the other hand the HTML 4 spec[1] reads as though only options and not option groups can be selected, though it does not say so explicitly. What should Dillo do about this? [1] http://www.htmlhelp.com/reference/html40/forms/select.html

1 0

patch: dialog.cc window deletion
by corvid＠lavabit.com May 5, 2008

May 5, 2008

Using the crude tool of top(1), I learned that viewing the source of a 500K file meant losing 500K down the drain. - If I give the text to the TextDisplay's initial buffer, it knows that it owns it. - Window::destroy() doesn't delete. The header comment in fltk/src/Window.cxx says it's some sort of 'hide' function: "Hides the window and also deletes all window system information about the window, and thus returns it back to the state it was in before the first show(). It is harmless to call this if the window is already destroyed." - And while I was in there, I threw in some callback(window_close_cb, ...)s for closing windows using the window mgr instead of a button.

2 1

[patch]: FltkListResource::sizeRequest()
by onepoint＠starurchin.org May 5, 2008

May 5, 2008

Currently FltkListResource::sizeRequest() is a stub with some hardwired values that are not very useful. This patch replaces it with something more appropriate (closely following the existing implementation of FltkOptionMenuResource::sizeRequest() ). Comments are welcome, Jeremy Henty

1 0

[BUG]: html.cc: broken Dlist iteration
by onepoint＠starurchin.org May 5, 2008

May 5, 2008

Html_build_query_data() and Html_make_multipart_boundary() iterate over a Dlist while destructively modifying it *and* recalculating the list length each time, so the iteration terminates prematurely. This makes Html_build_query_data() mangle the result when an input returns multiple values: half of the values get attached to the next input! This patch fixes the problem by caching the length of the list before modifying it so that the entire list is processed. Regards, Jeremy Henty

2 1

[patch]: select multiple
by onepoint＠starurchin.org May 5, 2008

May 5, 2008

This patch implements the MULTIPLE attribute for SELECT elements. It's small but it depends on the various patches and bugfixes I have already posted. Please comment. Regards, Jeremy Henty

1 0

[BUG]: lout::misc::StringBuffer::clear() does not invalidate the string
by onepoint＠starurchin.org May 3, 2008

May 3, 2008

Well, *that* was a bit of a wild goose chase! Remember I complained of selection problems in ListResources? It was nothing of the kind. Instead test/dw-ui-test was spuriously reporting that things were still selected when everything was in fact unselected. And the cause was a bug in lout::misc::Stringbuffer ! lout::misc::StringBuffer::clear() omits to invalidate the internal cached string, so if you call lout::misc::StringBuffer::getChars() (which updates the cached string and marks it as valid), then lout::misc::StringBuffer::clear() and then *immediately* call lout::misc::StringBuffer::getChars() again, you get the contents of the StringBuffer before you cleared it (because the cached value has not been invalidated), instead of the empty string. D'oh! One-liner patch attached. Regards, Jeremy Henty

3 4

dillo-fltk on x86_64
by 4winter＠informatik.uni-hamburg.de May 2, 2008

May 2, 2008

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hey@all :) I ran into some issues compiling dillo. I solved / worked around them, patches are attached. Here are my findings and my interpretations (let me know in case I am way off somewhere): dw2 === gcc-4.2 on linux/x86_64 complained about two casts in lout/object.cc: object.cc:118: error: cast from ?void*? to ?int? loses precision I felt it was nonsense to check the size of void* at runtime and used the preprocessor to choose which instruction to use based on the value of SIZEOF_VOID_P. dillo2 ====== There was another complaint by gcc: dialog.cc:249: error: cast from ?void*? to ?int? loses precision A parameter of type void* for a callback is used to carry a small integer (most likely the number of the button pressed or something). I solved this one by first casting to long and then to int. I saw this somewhere on the net, don't know if this is too hacky. dillo2/dpi ========== downloads.cc failed to compile since gcc couldn't find the fltk includes (I chose not to install them to /usr/local/lib), so I poked around in dpi/Makefile.am and managed to include LIBFLTK_CXXFLAGS into cxxflags (/me is an automake n00b). With my three patches applied dillo compiles and runs fine, however I had to create ~/.dillo and manually start dpid. Did I miss something here? Cheers, Justus - -- gpg key fingerprint: C82D 382A AB38 1A54 5290 19D6 A0F9 B035 686C 6996 -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFID6/ToPmwNWhsaZYRApgwAJsENlC0vsniya14RBTU8yrXVVjVyACff2HD GLM/iKxXf3UjBA6+GFOSE78= =JuDE -----END PGP SIGNATURE-----

5 19

[patch]: take 2: FtlkListResource
by onepoint＠starurchin.org May 1, 2008

May 1, 2008

Since Johannes' criticisms were (IIRC) mostly about my patch's complexity I hope he will look at a new patch that sticks to my original approach but is a lot simpler. I realised that if the user_data() of the Menu Widget items stored the index of the Item instead of the Item itself then FltkListResource could store the selections by index in a misc::SimpleVector<bool>. This means there is no need to wrap bools in object::Object pointers and no need to add instance methods to Items to make them Hashable. Most of the scaffolding of the previous patch just disappears (the line count decreases by a third). I don't want to dismiss Johannes' alternative suggestion out of hand, but I still think my approach has a lot going for it and I'd like people to have a chance to criticise my best shot at it. Comments are welcome! Jeremy Henty

2 1

[patch]: FtlkListResource select callbacks
by onepoint＠starurchin.org May 1, 2008

May 1, 2008

This is a first cut at making FltkListResource work properly. The resource maintains a hash of its items' selected status, the MultiBrowser widget gets a callback that updates the hash and isSelected() looks up the hash for the answer. It's the same technique I used for FltkOptionMenuResource , where the FLTK widgets are linked to their resources by user_data() and callback arguments. I added equals() and hashValue() methods to Items so they could be hash keys, and implemented an object::Object wrapper class for boolean values to be the hash values. The wrapper class is inside FltkListResource , though it could easily move to lout::object IMHO. The patch also enables the ListResource widget in test/dw-ui-test . Unfortunately it's still a little buggy. Specifying multiple selected items initially doesn't work right: they are all selected in the resource but only one is selected on the FLTK widget. Things also go out of sync when you select a group item: any previously selected item is unselected in the widget but not in the resource. I'll look into these further. Comments are welcome! Regards, Jeremy Henty

2 2

[patch] ignore SIGPIPE
by Johannes.Hofmann＠gmx.de April 30, 2008

April 30, 2008

Hello, this patch avoids exits on SIGPIPE. Currently this is only observed on DragonFly, but dillo1 had this too. Cheers, Johannes

2 1

[patch] cleanup iterator fix
by Johannes.Hofmann＠gmx.de April 30, 2008

April 30, 2008

Hi, now I noticed, that there is a more elegant way to create an iterator that is winded forward until its end. This restores symmetry between DeepIterator::next() and DeepIterator::prev(). Cheers, Johannes

2 1

[patch] two more tiny memory leaks
by Johannes.Hofmann＠gmx.de April 30, 2008

April 30, 2008

Hi, two more small leaks in dw2 found by valgrind. Cheers, Johannes

2 1

start on image maps (dw-part)
by Johannes.Hofmann＠gmx.de April 30, 2008

April 30, 2008

Hi, just to let you know, I just started with client side image maps (the dw-part). Below is an initial implementation of ImageMapsList. This is not yet ready at all, but perhaps someone might want have a look. Cheers, Johannes diff -r 877bbc03146b -r ebc6de688732 dw/image.cc --- a/dw/image.cc Thu Apr 17 22:14:42 2008 +0200 +++ b/dw/image.cc Mon Apr 21 21:03:51 2008 +0200 @@ -25,12 +25,44 @@ namespace dw { +ImageMapsList::ImageMap::ImageMap () +{ + shapesAndLinks = new container::typed::List <ShapeAndLink> (true); +} + +void ImageMapsList::ImageMap::add (core::Shape *shape, int link) { + ShapeAndLink *shapeAndLink = new ShapeAndLink (); + shapeAndLink->shape = shape; + shapeAndLink->link = link; + shapesAndLinks->append (shapeAndLink); +} + +int ImageMapsList::ImageMap::link (int x, int y) { + container::typed::Iterator <ShapeAndLink> it; + int link = -1; + + for (it = shapesAndLinks->iterator (); it.hasNext (); ) { + ShapeAndLink *shapeAndLink = it.getNext (); + + if (shapeAndLink->shape->isPointWithin (x, y)) { + link = shapeAndLink->link; + break; + } + } + + return link; +} + ImageMapsList::ImageMapsList () { + imageMaps = new container::typed::HashTable <object::Object, ImageMap> + (true, true); + currentMap = NULL; } ImageMapsList::~ImageMapsList () { + delete imageMaps; } /** @@ -42,6 +74,8 @@ ImageMapsList::~ImageMapsList () */ void ImageMapsList::startNewMap (object::Object *key) { + currentMap = new ImageMap (); + imageMaps->put (key, currentMap); } /** @@ -52,6 +86,7 @@ void ImageMapsList::startNewMap (object: */ void ImageMapsList::addShapeToCurrentMap (core::Shape *shape, int link) { + currentMap->add (shape, link); } // ---------------------------------------------------------------------- diff -r 877bbc03146b -r ebc6de688732 dw/image.hh --- a/dw/image.hh Thu Apr 17 22:14:42 2008 +0200 +++ b/dw/image.hh Mon Apr 21 21:03:51 2008 +0200 @@ -22,6 +22,27 @@ namespace dw { */ class ImageMapsList { +private: + class ImageMap: public object::Object { + private: + class ShapeAndLink: public object::Object { + public: + core::Shape *shape; + int link; + }; + + container::typed::List <ShapeAndLink> *shapesAndLinks; + public: + ImageMap (); + ~ImageMap (); + + void add (core::Shape *shape, int link); + int link (int x, int y); + }; + + container::typed::HashTable <object::Object, ImageMap> *imageMaps; + ImageMap *currentMap; + public: ImageMapsList (); ~ImageMapsList ();

3 20

patch: motionNotify and LEAVE
by corvid＠lavabit.com April 30, 2008

April 30, 2008

Since findtext has been in the news lately, I might as well at least provide this bit so that my code behaves better with dialogs.

2 3

[PATCH]: call widget->layout_damage() from FltkWidgetView::allocateFltkWidget
by onepoint＠starurchin.org April 30, 2008

April 30, 2008

This patch makes FltkWidgetView::allocateFltkWidget mark the FLTK widget's layout as damaged. This fixes dw::fltk::ui::FltkListResource (which is necessary to implement MULTIPLE SELECT inputs in forms). Currently a dw::fltk::ui::FltkListResource widget fails to display itself. (You can run test/dw-resource-test to see this). This is because the resource's fltk::MultiBrowser widget is created with a tiny size and later reallocated to the correct size. The viewport calls layout() each time, but this doesn't call widget->layout() unless the widget's layout is damaged. Unfortunately the layout is marked as damaged when the widget is created but *not* after it is resized, so MultiBrowser::layout() gets called when the widget is tiny, but not when it is the right size. Alas, MultiBrowser::draw() uses size information that was cached by MultiBrowser::layout() , so it behaves as though the widget is still tiny, concludes that none of the Items are visible and doesn't draw anything. This bug potentially bites all widgets, it's just sheer luck that only MultiBrowser is visibly affected. Other widgets are either simple enough to survive their layout() being called when they were a different size, or else smart enough to mark *themselves* as damaged so that the viewport calls their layout() at the right time. Regards, Jeremy Henty

2 3

patch: usemap <area shape=default>
by corvid＠lavabit.com April 30, 2008

April 30, 2008

shape=default is the background in a client-side image map.

2 1

Q: lout::Object::Bool ?
by onepoint＠starurchin.org April 30, 2008

April 30, 2008

I'm looking at properly implementing dw::fltk::ui::FltkListResource::isSelected() and I need to maintain a hash of Items to boolean values. However, lout::container::typed::HashTable requires the keys and values to inherit lout::object::Object . I could use lout::object::Integer values, but the really appropriate thing would be lout::object::Bool , which doesn't exist. Is it worth implementing lout::object::Bool or should I just use lout::object::Integer ? Regards, Jeremy Henty

1 0

[patch]: remove redundant #include in fltkflatview.cc
by onepoint＠starurchin.org April 30, 2008

April 30, 2008

A small cleanup: the "#include <fltk/damage.h>" in fltkflatview.cc is unnecessary. Regards, Jeremy Henty

2 1

[patch] fix table destructor
by Johannes.Hofmann＠gmx.de April 30, 2008

April 30, 2008

Hi, with memleaks in the news lately.... Here comes another one. Cheers, Johannes

2 1

patch: fltkimgbuf memleak
by corvid＠lavabit.com April 29, 2008

April 29, 2008

You should see a significant decrease in memory used.

3 2

Cheering some recent improvements in table layout
by onepoint＠starurchin.org April 29, 2008

April 29, 2008

Just to say that I'm seeing much better table rendering now. I have a page with a large two-column table. For a while recently this rendered very strangely in Dillo: it appeared with the two columns at opposite sides of the page, then the second column would move over to the left, one item at a time, to sit next to the first. But now the table immediately appears with both columns in the right place. I don't know what caused the strange realignment behaviour but I'm glad it's gone. Regards, Jeremy Henty

2 1

[patch] phrase search - take 2
by Johannes.Hofmann＠gmx.de April 29, 2008

April 29, 2008

Hi, here comes my second attempt on phrase search. As already described, I changed CharIterator back to not return \0s. Instead blanks are inserted when the space flag is set. This way CharIterator always returns a smooth stream of chars whether for normal html- or pre-text. Then there was a bug in DeepIterator::prev() that took me some time to find. I also adjusted charsEqual() to treat all space-chars equally. Finally I had to add a check in CharIterator::highlight() to avoid a crash when searching for the last word in pre-text with a blank appended. Cheers, Johannes

2 1

dillo seems to close file descriptors twice
by Johannes.Hofmann＠gmx.de April 29, 2008

April 29, 2008

Hi, I just noticed, that dillo seems to close file descriptors twice via IO_close_fd, which causes problems when other threads run in parallel for name server queries. Here are the two call stacks via which IO_close_fd is called for each fd: (gdb) b IO_close_fd Breakpoint 1 at 0x8070781: file IO.c, line 137. (gdb) r dillo.org Starting program: /home/hofmann/projects/dillo/plain/dillo2-hg/src/dillo-fltk dillo.org dillo_dns_init: Here we go! (threaded) Enabling cookies as from cookiesrc... Nav_open_url: new url='http://dillo.org' Dns_server [0]: dillo.org is 0x28559680 Connecting to 134.102.206.165 Nav_open_url: new url='http://www.dillo.org/' Breakpoint 1, IO_close_fd (io=0x28550a00, CloseCode=2) at IO.c:137 137 fprintf(stderr,"====> begin IO close (%d)\n", io->FD); Current language: auto; currently c (gdb) bt #0 IO_close_fd (io=0x28550a00, CloseCode=2) at IO.c:137 #1 0x0807094e in a_IO_ccc (Op=4, Branch=1, Dir=2, Info=0x285509e0, Data1=0x0, Data2=0x0) at IO.c:392 #2 0x0805633e in a_Chain_bcb (Op=4, Info=0x28550a00, Data1=0x0, Data2=0x0) at chain.c:98 #3 0x0806ea14 in a_Http_ccc (Op=4, Branch=1, Dir=2, Info=0x2854df60, Data1=0x0, Data2=0x0) at http.c:522 #4 0x0805633e in a_Chain_bcb (Op=4, Info=0x28550a00, Data1=0x0, Data2=0x0) at chain.c:98 #5 0x0805a514 in a_Capi_ccc (Op=4, Branch=1, Dir=2, Info=0x2854df40, Data1=0x0, Data2=0x0) at capi.c:491 #6 0x0805a39c in a_Capi_ccc (Op=4, Branch=2, Dir=1, Info=0x285509a0, Data1=0x0, Data2=0x0) at capi.c:601 #7 0x0805630a in a_Chain_fcb (Op=4, Info=0x28550a00, Data1=0x0, Data2=0x0) at chain.c:86 #8 0x0806ffd1 in a_Dpi_ccc (Op=4, Branch=2, Dir=1, Info=0x28550bc0, Data1=0x0, Data2=0x0) at dpi.c:671 #9 0x0805630a in a_Chain_fcb (Op=4, Info=0x28550a00, Data1=0x0, Data2=0x0) at chain.c:86 #10 0x08070891 in a_IO_ccc (Op=4, Branch=2, Dir=1, Info=0x28550c00, Data1=0x28550c20, Data2=0x0) at IO.c :410 #11 0x08070b64 in IO_callback (fd=<value optimized out>, io=0x28550c20) at IO.c:201 #12 0x08070beb in IO_fd_read_cb (fd=7, data=0x2) at IO.c:277 #13 0x080bcfe5 in fltk::wait () #14 0x080bd195 in fltk::run () #15 0x0804eadb in main (argc=0, argv=0x0) at dillo.cc:110 (gdb) c Continuing. ====> begin IO close (7) Breakpoint 1, IO_close_fd (io=0x28550c20, CloseCode=2) at IO.c:137 137 fprintf(stderr,"====> begin IO close (%d)\n", io->FD); (gdb) bt #0 IO_close_fd (io=0x28550c20, CloseCode=2) at IO.c:137 #1 0x0807089e in a_IO_ccc (Op=4, Branch=2, Dir=1, Info=0x28550c00, Data1=0x28550c20, Data2=0x0) at IO.c:411 #2 0x08070b64 in IO_callback (fd=<value optimized out>, io=0x28550c20) at IO.c:201 #3 0x08070beb in IO_fd_read_cb (fd=7, data=0x2) at IO.c:277 #4 0x080bcfe5 in fltk::wait () #5 0x080bd195 in fltk::run () #6 0x0804eadb in main (argc=0, argv=0x0) at dillo.cc:110 (gdb) c Continuing. ====> begin IO close (7) A line like fprintf(stderr,"====> begin IO close (%d)\n", io->FD); at the beginning of IO_close_fd() also shows the problem. Cheers, Johannes

3 3

Since we've been talking about name clashes, what about "assert"
by onepoint＠starurchin.org April 27, 2008

April 27, 2008

Because lout::misc defines an inline assert function, including dw/core.hh fails if you've already included anything that defines the standard assert macro (such the C++ iostream stuff). You can work around it by including dw/core.hh first, but that feels pretty fragile as it depends on nothing subsequently using the standard assert. What's the best fix? (Example appended.) Regards, Jeremy Henty For example: #include <ios> #include <dw/core.hh> yields (with GCC 3.4.3): In file included from ./../Source/current/dw2/dw/../lout/object.hh:7, from ./../Source/current/dw2/dw/core.hh:39, from dw_assert.cc:9: ./../Source/current/dw2/dw/../lout/misc.hh:36:55: macro "assert" passed 3 arguments, but takes just 1 In file included from ./../Source/current/dw2/dw/../lout/object.hh:7, from ./../Source/current/dw2/dw/core.hh:39, from dw_assert.cc:9: ./../Source/current/dw2/dw/../lout/misc.hh:37: error: invalid function declaration ./../Source/current/dw2/dw/../lout/misc.hh:185:30: macro "assert" passed 4 arguments, but takes just 1 ./../Source/current/dw2/dw/../lout/misc.hh:198:30: macro "assert" passed 4 arguments, but takes just 1 ./../Source/current/dw2/dw/../lout/misc.hh:213:30: macro "assert" passed 4 arguments, but takes just 1 ./../Source/current/dw2/dw/../lout/misc.hh: In member function `T* lout::misc::SimpleVector<T>::getRef(int)': ./../Source/current/dw2/dw/../lout/misc.hh:184: error: `assert' undeclared (first use this function) ./../Source/current/dw2/dw/../lout/misc.hh:184: error: (Each undeclared identifier is reported only once for each function it appears in.) ./../Source/current/dw2/dw/../lout/misc.hh: In member function `T lout::misc::SimpleVector<T>::get(int)': ./../Source/current/dw2/dw/../lout/misc.hh:197: error: `assert' undeclared (first use this function) ./../Source/current/dw2/dw/../lout/misc.hh: In member function `void lout::misc::SimpleVector<T>::set(int, T)': ./../Source/current/dw2/dw/../lout/misc.hh:212: error: `assert' undeclared (first use this function) make[1]: *** [dw_assert-dw_assert.o] Error 1

2 2

Re: patch: various IN_*s
by jcid＠dillo.org April 27, 2008

April 27, 2008

Hi, On Sun, Apr 27, 2008 at 02:15:37PM +0000, corvid wrote: > - disallow <button> or <input> inside <button>. > Nested buttons and image inputs inside buttons don't come out right, > and they aren't legal anyway. OK. > - disallow <input> inside <select>. > As I mentioned before, </select> does not like that, and it isn't legal. ok. > - disallow <isindex> inside <form>. > It's legal (%block goes inside form, and isindex is a %block), > but it splits a form into pieces, given dillo's current implementation. > Firefox handles it as a form within a form, but I don't want to add > complexity, and I know you don't want to add complexity :) ok. Committed. -- Cheers Jorge.-

1 0

"View Page source" scrolling problem
by 4winter＠informatik.uni-hamburg.de April 27, 2008

April 27, 2008

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hey folks :) I found another bug / feature... I use a trackstick with mouse wheel emulation. This allows me to scroll text by pressing the middle button and moving the stick. Someone in the xorg input layer maps this to buttons 4 - 7 (up, down, left, right). Problem is that it is impossible to just scroll up. You end up being off by some degree and you get a event for left and right every now and then (confirmed using xev). Unfortunately this kills the scrolling in the view page source window: http://teythoon.cryptobitch.de/stuff/scroll-problem.ogg It would also be nice to be able to scroll left/right with buttons 6 and 7. Since there are mouses that have special navigation buttons (most likely also being number 6 and 7) it would be nice to be able to configure the mapping (firefox does this) for these buttons. That would make trackstick users happy and also those of us that are stuck with a mouse with those navigation buttons (I have access to one of those and the buttons are located at the left side so that you can press them with your thumb [not a very left-handed friendly design]. To make a long story short you just end up pressing them by accident the moment you filled out some huge form...). But then again, some people might find them useful. Cheers, Justus - -- gpg key fingerprint: C82D 382A AB38 1A54 5290 19D6 A0F9 B035 686C 6996 -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFIE+EdoPmwNWhsaZYRArv+AJ0ZO6hR5Caw/eaHXtnwWv9RFEgETACfYn6v FcJ5rxwibEXoAK5ZLtmTNEs= =t0nP -----END PGP SIGNATURE-----

1 0

[PATCH]: lout::misc::min
by onepoint＠starurchin.org April 27, 2008

April 27, 2008

I mean, this *has* to be a typo, right? Jeremy Henty

2 3

Docs question: dw::core::ui::ListResource::SelectionMode::SELECTION_AT_MOST_ONE
by onepoint＠starurchin.org April 27, 2008

April 27, 2008

The description of dw::core::ui::ListResource::SelectionMode::SELECTION_AT_MOST_ONE does not make sense. Should "The user may not the only selected item." read "The user may unselect the only selected item."? Regards, Jeremy Henty

3 4

[patch] phrase search
by Johannes.Hofmann＠gmx.de April 26, 2008

April 26, 2008

Hi, attached is a patch that make phrase search work. I'm not 100% sure about the CharIterator::prev() change. CharIterator::next() returned \0 between words, prev() didn't. So something was wrong. I decided to go with current next() behaviour and made prev() return \0 as word separator too. Also the current implementation only works if one enters exactly one blank between words in the search dialog. Should we collapse multiple blanks into one automatically? Cheers, Johannes

2 2

Re: patch: Accept-Charset: utf-8,*;q=0.8
by jcid＠dillo.org April 26, 2008

April 26, 2008

On Thu, Apr 17, 2008 at 04:57:48PM +0000, corvid wrote: > It is good for dillo to state its preference for UTF-8 > over ISO-8859-1. > > Also, the RFC says that not mentioning '*' puts dillo > in theoretical danger of receiving 406 Not Acceptable. > Committed, assuming: "Accept-Charset: utf-8,*;q=0.8\r\n" means dillo2 leaves the door open to also receive iso-8859-1. -- Cheers Jorge.- > --- dillo2/src/IO/http.c 2008-04-11 19:48:51.000000000 +0000 > +++ dillo2-cur/src/IO/http.c 2008-04-17 16:17:17.000000000 +0000 > @@ -236,7 +236,7 @@ Dstr *a_Http_make_query_str(const DilloU > query, > "POST %s HTTP/1.1\r\n" > "Connection: close\r\n" > - "Accept-Charset: utf-8, iso-8859-1\r\n" > + "Accept-Charset: utf-8,*;q=0.8\r\n" > "Accept-Encoding: gzip\r\n" > "Host: %s%s\r\n" > "%s" > @@ -258,7 +258,7 @@ Dstr *a_Http_make_query_str(const DilloU > "GET %s HTTP/1.1\r\n" > "%s" > "Connection: close\r\n" > - "Accept-Charset: utf-8, iso-8859-1\r\n" > + "Accept-Charset: utf-8,*;q=0.8\r\n" > "Accept-Encoding: gzip\r\n" > "Host: %s%s\r\n" > "%s"

2 2

patch: /* todo: don't create things if button type is unknown */
by corvid＠lavabit.com April 26, 2008

April 26, 2008

2 1

Re: the last patch that I sent you BREAKS isindex.
by jcid＠dillo.org April 26, 2008

April 26, 2008

On Thu, Apr 17, 2008 at 06:17:04AM +0000, corvid wrote: > I shouldn't have moved isindex's connectActivate to close_form. > If you wish, I can make a new version. Yes, please. > > By the way, I found that the top of open_input needs > if (html->InFlags & IN_SELECT) { > MSG_HTML("INPUT inside SELECT\n"); > return; > } > Otherwise, if another input sneaks in there, close_select will crash. Please include it in the patch. -- Cheers Jorge.-

2 1

[patch] introduce lout namespace (was: namespace problem)
by Johannes.Hofmann＠gmx.de April 26, 2008

April 26, 2008

Hi, while trying to implement a fix to the namespace problem I came to the conclusion that the easiest and least intrusive way is to keep the existing namespaces in the lout directory and just wrap them in another namespace "lout". This fixes the name collision with signal. Cheers, Johannes

2 1

I figured out why charset handling has been all broken lately.
by corvid＠lavabit.com April 26, 2008

April 26, 2008

The top of Nav_open_url has bool_t ForceReload = (URL_FLAGS(url) & (URL_E2EReload + URL_ReloadFromCache)); When repushing URLs, URL_FLAGS contains URL_ReloadFromCache, which is 0x100. And remember how Jorge changed bool_t from int to unsigned char? I'd forgotten about it, too, which meant I wasted some time looking in the wrong places for this bug.

3 2

patch: image loading on demand for complex button
by corvid＠lavabit.com April 26, 2008

April 26, 2008

(Jorge: Okay. That gives me an opportunity to put this in the priority queue instead.) I was able to get the menus to work for buttons and image inputs, so image loading can be conditional now. My usual disclaimer applies regarding the dw2 side that I did something that made it work, but it may not be the right way.

2 1

namespace problem
by Johannes.Hofmann＠gmx.de April 24, 2008

April 24, 2008

Hi, some time ago I already reported a compilation problem that currently only exists on DragonFlyBSD. The namespace "signal" in the lout library collides with a function declaration in /usr/include/signal.h. For some reason DragonFly happens to include signal.h indirectly. But the same problem will appear on Linux too once we should need signal.h somewhere. Just try to compile the following C++ file (g++ -c <file>): #include <signal.h> namespace signal { } I'd like to resolve this issue before release. I see two options: * rename namespace "signal" to "loutsignal" - which is pretty ugly. The concept of namespaces exists to avoid such stuff... * put all classes in the lout directory into a single namespace "lout". At the moment I prefer the second option, but I don't know Sebastian's reasoning behind this naming. Any opinions? Cheers, Johannes

2 1

dillo-fltk exits on SIGPIPE
by Johannes.Hofmann＠gmx.de April 23, 2008

April 23, 2008

Hi, do you also see exits on SIGPIPE e.g. on: http://daily.daemonnews.org dillo 1 had signal(SIGPIPE, SIG_IGN); which dillo-fltk is missing. Cheers, Johannes

3 2

Good news from fltk!
by Johannes.Hofmann＠gmx.de April 19, 2008

April 19, 2008

Hi, I stumbled across this post on fltk.bugs: http://fltk.org/newsgroups.php?s5281+gfltk.bugs+v5290+T0 This is great for us, as it allows us to implement efficient double buffering without X11 specific hacks. Attached is a patch that enables double buffering for partial screen redraws (this is where the flickering is most annoying). You will need at least version r6079 of fltk2 to test. Cheers, Johannes

4 27

Collapsing whitespace
by jcid＠dillo.org April 17, 2008

April 17, 2008

Hi, Just submitted a patch I wanted to make long ago. It deals with collapsing whitespace. Here's one example: http://www.w3.org/TR/REC-html40/intro/sgmltut.html#h-3.2.1 BTW, I also changed the bool_t to "unsigned char". Please test whether it compiles cleanly. -- Cheers Jorge.-

2 1

Re: patch: BUTTON
by jcid＠dillo.org April 15, 2008

April 15, 2008

On Tue, Apr 15, 2008 at 04:16:55AM +0000, corvid wrote: > The old code showed uncertainty about whether one can interact with the > stuff inside a button. The spec doesn't say. I suspect you aren't > supposed to be able to. Firefox doesn't let you. If FF doesn't allow interaction it will be hardly found in the WEB, unless IE allows. :-P > When trying to look up info on button nesting, I discovered -- by crashing -- > that I needed to test whether I was in a form. However, I learned that > section 17.2.1 says: > > "The elements used to create controls generally appear inside a FORM element, > but may also appear outside of a FORM element declaration when they are used > to build user interfaces. This is discussed in the section on intrinsic > events. [p.254] Note that controls outside a form cannot be successful > controls [p.245]." > > ..but I'll make a separate patch for controls outside of forms. AFAICS this may be useful with client side scripts, but we're a long way from suporting scripts. -- Cheers Jorge.-

2 1

[patch] fix cookies freezing dillo problem
by akemnade＠tzi.de April 15, 2008

April 15, 2008

Hi, after some debugging I found out that dillo was frozen because it waits for input on the unix socket from the cookies.dpi. cookies.dpi itself blocks on the read() on the connection socket. If you kill dillo and stop dpid, the next time the cookies.dpi starts, it finds the cookies files locked and terminates immedediately. Then dpid starts the cookies.dpi again and again in an infinite loop. In my theory, the freeze where both processes are hanging in read() is caused when a connection is creating via DEBUG_MSG(5, "Cookies.c: a_Cookies_set \n\t \"%s\" \n",cmd ); a_Capi_dpi_send_cmd(NULL, NULL, cmd, "cookies", 1); but the data is not send yet and /* Get the answer from cookies.dpi */ dpip_tag = a_Dpi_send_blocking_cmd("cookies", cmd); is called. I have attached a patch which uses a_Dpi_send_blocking_cmd(). It seems to fix the problem. dillo has not been frozen again yet. Maybe cookies.dpi should create the SockHandler only when there is data to read (checking for it with select() first). Greetings Andreas Kemnade

2 3

patch: image inputs
by corvid＠lavabit.com April 15, 2008

April 15, 2008

It works on the sites I tried, and the data sent looks right for other cases, but: - To get click position, I used event_x() and event_y() in fltkui.cc because they happened to have the right values in them and I didn't know the proper place to look for this information. - If the image could not be loaded and you click on the alt text, the click position is sent, even though it's not meaningful. - I don't prevent the image from being loaded even when automatic image loading is disabled (but I'm the only person who would notice anyway). (I would have to be able to get button 3 to bring up the image menu.)

2 5

Re: patch: file input
by jcid＠dillo.org April 15, 2008

April 15, 2008

Hi, On Thu, Apr 10, 2008 at 04:13:56PM +0000, corvid wrote: > I just stuck it on a label button for now. It looks quite informative. > Already an improvement over Firefox, where I only > find out that I mistyped the filename when the server > doesn't get the file. > > Somewhere that accepts files: http://validator.w3.org > > Edge case that I just noticed: > If I read in a file, delete that file, and try to read it in again > (uses E2EReload), the submit will send the (now empty) file because > a_Capi_get_buf() finds the (now empty) entry in cache. > So I'll have to find out where an a_Cache_entry_remove_by_url() should go. Yes. Please also add a file size threshold, to avoid a dillo exit from an out of memory condition. Committed! -- Cheers Jorge.-

2 10

[patch] fix warning in html.cc
by Johannes.Hofmann＠gmx.de April 14, 2008

April 14, 2008

Hi, html.cc: In function 'void a_Html_form_event_handler(void*, form::Form*, void*)': html.cc:4331: warning: 'input' may be used uninitialized in this function can be fixed with attached patch. Please review. Cheers, Johannes

3 2