- Dillo-dev - lists.mailman3.com

Work on TABLES

by jcid＠dillo.org

On Sat, Jan 19, 2008 at 07:19:13PM +0000, place wrote: > table.cc is intimidating stuff. > > It works so far, so I don't have much to report, "It works so far" is great news for me, and certainly if it keeps stable better. Good feedback is welcomed! > but I just needed to say > that when I saw per.bak.html, I thought it was going to be something derived > from some article on http://en.wikipedia.org/wiki/Per_Bak instead of being > about, you know, percentages. Percentages is definitely less cool than frozen insulators... -- Cheers Jorge.-

Jan. 19, 2008

Re: patch: image.cc typo one-liner

by place＠gobigwest.com

I wrote: > Ah, the ancient programming principle that > The one who touched it last owns it. It dawned on me that I just touched textblock.cc. Oh no! *runs screaming*

Jan. 19, 2008

Various minor patches

by jcid＠dillo.org

On Sat, Jan 19, 2008 at 05:23:50PM +0100, Thomas-Martin Seck wrote: > * Thomas-Martin Seck (tmseck-lists(a)netcologne.de) > > > Sorry, I had hoped that the autotools could deal with this somehow and > > since I only use FreeBSD I could only get to the "works for me" state. > > What does break on Linux? The intent is to check for libiconv.h in the > > default include path and for iconv_open in libiconv, assuming that > > libiconv can be located without adding -L/path/to/libiconv to LDFLAGS > > explicitly. > > OK, I learned that glibc(?) nowadays includes iconv functionality. I > updated the patch so that it checks for iconv_open in libc first and, > failing this, in libiconv next. It still works on FreeBSD 6.2, could you > please test whether this is the correct approach for Linux et al? It works here, committed! Comments welcome. -- Cheers Jorge.-

Jan. 19, 2008

Re: patch: image.cc typo one-liner

by place＠gobigwest.com

Jorge wrote: > Now that you're images, would you mind giving a look to these anomalies? Ah, the ancient programming principle that The one who touched it last owns it. I have met with this before :) And a sobering thought, now, glancing at the changelog... > > http://www.dillo.org/test/img/ > > 1.- n1 renders a right column with old background (over dialup). I couldn't tell what this one was supposed to be doing wrong. > 2.- n2 makes space inside the link, not outside, and renders different > image size the first time (e.g. when from local harddisk); after > back and forward size changes. > 3.- n3 is obviously wrong! ;) I don't really know what is going on with allocation and requisition and all of this drawing business, but I did find that after - requisition->width += getStyle()->boxDiffHeight (); + requisition->width += getStyle()->boxDiffWidth (); in my Image::sizeRequestImpl -- which didn't seem to change anything, but sure _sounds_ sensible, and diff -pur dw2/dw/textblock.cc dw2-cur/dw/textblock.cc --- dw2/dw/textblock.cc 2007-12-20 20:02:20.000000000 +0000 +++ dw2-cur/dw/textblock.cc 2008-01-19 18:31:26.000000000 +0000 @@ -339,12 +339,12 @@ void Textblock::sizeAllocateImpl (core:: childAllocation.y = lineYOffsetCanvasAllocation (line, allocation) + (line->ascent - word->size.ascent) - - word->content.widget->getStyle()->margin.top; +; // - word->content.widget->getStyle()->margin.top; childAllocation.width = word->size.width; childAllocation.ascent = word->size.ascent - + word->content.widget->getStyle()->margin.top; +; // + word->content.widget->getStyle()->margin.top; childAllocation.descent = word->size.descent - + word->content.widget->getStyle()->margin.bottom; +; // + word->content.widget->getStyle()->margin.bottom; word->content.widget->sizeAllocate (&childAllocation); break; the problem went away for those cases, at least. But as for whether it's supposed to be like this or the margins should be included and something else is supposed to be subtracting them somewhere, I have no idea. I was led in this direction by finding that allocation.y was -95 somewhere in image.cc with n3, which just sounded wrong.

Jan. 19, 2008

Work on TABLES

by Johannes.Hofmann＠gmx.de

Hi Jorge, works great so far! Your patch makes www.wetter-online.de work again. It did work in dillo1, but didn't in dillo2 until now. It still renders a bit different than in firefox, but it is already better than in dillo1. I will continue testing. Enjoy your free time, Johannes On Sat, Jan 19, 2008 at 11:13:22AM -0300, Jorge Arellano Cid wrote: > Hi there, > > Good news! > > It looks like awakening the dragon wasn't a bad idea after all! > > It took a lot of time and debugging, but now after lots of tests, > it seems much better than what we had before. > > This is not a simple patch, but an algorithms change + bug fixes. > Beware that the html documentation is not up-to-date yet though; > I prefer to let you test the new code with lots of other cases and > to send feedback or bugfixes to the list before adopting the new code. > > I uploaded a subset of the used test cases to: > > http://www.dillo.org/test/tables > > You can start with per1.html, per2.html per3.html & per4.html. > > Please note that the test is not just rendering the page, but resizing > its tables from MIN to MAX. It's interesting to compare what we had before, > dillo1 and Firefox (Ill cases follow Firefox's example). > > So the bottom line is: Test, test, test!!! & send feedback. > > I'll be back near Jan 30; after enjoying my deserved rest period. ;-) > > -- > Cheers > Jorge.- > > PS: If you want to debug, uncomment table.cc:21 > and define dillo_dbg_rendering to 1 in html.cc:64. > PS2: I left an exit assert testing d_a and d_w >= 0. If this triggers, > please send the URL. > > _______________________________________________ > Dillo-dev mailing list > Dillo-dev(a)dillo.org > http://lists.auriga.wearlab.de/cgi-bin/mailman/listinfo/dillo-dev

Jan. 19, 2008

Various minor patches

by tmseck-lists＠netcologne.de

* Thomas-Martin Seck (tmseck-lists(a)netcologne.de) > Sorry, I had hoped that the autotools could deal with this somehow and > since I only use FreeBSD I could only get to the "works for me" state. > What does break on Linux? The intent is to check for libiconv.h in the > default include path and for iconv_open in libiconv, assuming that > libiconv can be located without adding -L/path/to/libiconv to LDFLAGS > explicitly. OK, I learned that glibc(?) nowadays includes iconv functionality. I updated the patch so that it checks for iconv_open in libc first and, failing this, in libiconv next. It still works on FreeBSD 6.2, could you please test whether this is the correct approach for Linux et al?

Jan. 19, 2008

Re: patch: image.cc typo one-liner

by jcid＠dillo.org

Hi, On Sat, Jan 19, 2008 at 02:47:43AM +0000, place wrote: > It caused stripes along the sides of images with vspace and hspace. > > The line numbers of the patch are off because my image.cc is somewhat > hacked up, but who cares for a one-liner... > > diff -pur dw2/dw/image.cc dw2-cur/dw/image.cc > --- dw2/dw/image.cc 2007-12-24 06:17:14.000000000 +0000 > +++ dw2-cur/dw/image.cc 2008-01-19 02:18:15.000000000 +0000 > @@ -263,7 +263,7 @@ void Image::drawRow (int row) > > buffer->getRowArea (row, &area); > if (area.width && area.height) > - queueDrawArea (area.x + getStyle()->boxOffsetY (), > + queueDrawArea (area.x + getStyle()->boxOffsetX (), > area.y + getStyle()->boxOffsetY (), > area.width, area.height); > } Committed. Now that you're images, would you mind giving a look to these anomalies? http://www.dillo.org/test/img/ 1.- n1 renders a right column with old background (over dialup). 2.- n2 makes space inside the link, not outside, and renders different image size the first time (e.g. when from local harddisk); after back and forward size changes. 3.- n3 is obviously wrong! ;) -- Cheers Jorge.-

Jan. 19, 2008

Work on TABLES

by jcid＠dillo.org

Hi there, Good news! It looks like awakening the dragon wasn't a bad idea after all! It took a lot of time and debugging, but now after lots of tests, it seems much better than what we had before. This is not a simple patch, but an algorithms change + bug fixes. Beware that the html documentation is not up-to-date yet though; I prefer to let you test the new code with lots of other cases and to send feedback or bugfixes to the list before adopting the new code. I uploaded a subset of the used test cases to: http://www.dillo.org/test/tables You can start with per1.html, per2.html per3.html & per4.html. Please note that the test is not just rendering the page, but resizing its tables from MIN to MAX. It's interesting to compare what we had before, dillo1 and Firefox (Ill cases follow Firefox's example). So the bottom line is: Test, test, test!!! & send feedback. I'll be back near Jan 30; after enjoying my deserved rest period. ;-) -- Cheers Jorge.- PS: If you want to debug, uncomment table.cc:21 and define dillo_dbg_rendering to 1 in html.cc:64. PS2: I left an exit assert testing d_a and d_w >= 0. If this triggers, please send the URL.

Jan. 19, 2008

Working on TABLES

by jcid＠dillo.org

Hi there, As you may remember, I dared to awake the sleeping dragon of table apportion algorithms. Frankly, I don't yet know whether it was a good idea after all, but I have hope. It is impresive how complex the geometric combination of cases grows; hopefully some code will be uploaded before Jan 20. In the meanwhile, does anyone want to give a try to imagemaps? Or maybe to implement tabs? This is much simpler than it looks. Basically we can add the tab widgets and to switch the navigation stack and rendered page upon click. As dillo is quick to render this has the advantage of low memory consumption because we don't need to have all the rendered pages in memory! And it opens the possibility of multiple windows with TABS using low resources. -- Cheers Jorge.-

Jan. 16, 2008

patch: progressive images

by jcid＠dillo.org

Hi, On Mon, Jan 14, 2008 at 06:06:03PM +0000, place wrote: > (the ones that start out blurry and grow sharper) > > - Added code for progressive display of, well, progressive jpegs because > I was tired of waiting for full download before display. > jpeglib uses the word "scan" for each loop through the scanlines of the > image, so I did the same. I'm not sure it's a terribly descriptive term, > though, at least to me (not a graphics person). Couldn't think of a better > word. > > - Turned out scaled ones didn't work. (And, incidentally, that was frustrating > because it took some time to realize that the breaking ones were scaled > ever so slightly.) Added some bit clearing to imgbuf code on the dw side so > that scaled buffers would continue to copy from root buffers. > > - Turned out that if there was more than one copy on a page, only the first > copy worked. (dillo uses image decoders on the first one and pulls the > scanlines out of Dicache for the rest.) More bit clearing and a scan number > to help a_Dicache_callback with its copying. > > - Discovered that interlaced pngs use progressive display, and that they > hadn't been working for scaled or for multiple on a page. Great! Committed. > - Considered looking into animated gifs, but then my good sense prevailed :) :-) -- Cheers Jorge.-

Jan. 16, 2008