April 2009 - Dillo-dev - lists.mailman3.com

perhaps an iterator bug
by corvid＠lavabit.com 03 Sep '09

03 Sep '09

I was trying the new search functionality a bit more and found this (it's also present in my Dec 2008 dillo*). With the page: <table> <tr><td>no close-tr <tr><td>has a close-tr</tr> <tr><td>here's another line</tr> </table> if I search for "rh", it finds one instance of it split between the second and third table rows. * My older dillos were lost in a tragic "I don't think I need my CVS stuff anymore" accident.

3 4

Re: patch: media_type pref
by corvid＠lavabit.com 05 Aug '09

05 Aug '09

Jorge wrote: > On Mon, Feb 16, 2009 at 05:37:36AM +0000, corvid wrote: > > It makes more sense to have a pref that the user can > > control than for us to guess what they might like > > based on screen size or whatever, I think. > > I hesitate on this. For instance: if media_type is defined to > "handheld" and there's no such remote CSS, at least it should > fallback to "screen". If there is a handheld style sheet available and dillo is running on a handheld, it would be a shame not to use it. I'm hoping that the page author would have the sense to supply something for type "all". > Maybe it's more practical to provide custom stylesheets (as the > current "style.css". e.g. small_screen.css, print.css) and to > allow switching among them from the "Tools" menu. > > What do you think? I think someone using handheld would want to use both the page author's handheld-specific code and a user style suitable for their device. I'll send this to the list in case anyone with experience using such devices with the web has opinions to share.

3 3

[patch] Improving keyboard shortcuts for tab navigation
by tim.nieradzik＠gmx.de 13 Jun '09

13 Jun '09

This patch adds keyboard shortcuts for the tabs 0 - 9. You can reach them by pressing Ctrl concurrently with the number number of the tab you would like to see. Thus, for the first tab this is Ctrl-0. Alternatively, you can press Ctrl-Home. To get to the last tab, simply press Ctrl-End. In addition the modifier for moving left and right between tabs has been changed from Shift to Ctrl because it makes sense to have an unified modifier for all tab movement operations. --Tim

7 51

experimental patch: Re: dillo2 wrapping Chinese
by corvid＠lavabit.com 06 Jun '09

06 Jun '09

dillo people: It turns out that dillo handles Chinese quite badly because the text doesn't contain whitespace (BUG 827). It looks like Japanese is that way as well. 1) People using East Asian languages: Does this do roughly the right thing? I've read that there are picky rules about certain individual characters, but I just mean in general here. 2) Thoughts on optimization are welcomed. littlebat: For some reason, I was thinking that whitespace was required between words in dillo, but I misremembered. littlebat wrote: > Hi, > Thanks for the work on Dillo. > > I think it isn't a good idea to resolve long chinese line wrapping > problem by inserting space after punctuation mark("?" or "?" or "?", > etc..). > > Here is a "Dillo2 Wrap Long Chinese Line Test" page: > http://www.learndiary.com/en/dillo2-wrap-long-chinese-line-test.htm . > Open it with Firefox or IE, then open it with Dillo2, You will see the > different of wrapping long chinese line between them. > > Note, unlike english, there is no space between Chinese characters. So, > it should be wrapped after any Chinese character which has reached the > end of a line. I don't know the technique detail about this, but, I > think, let the Chinese characters function as (english) commas or > periods is a good idea, so, a new line can be started after any Chinese > charcter which character has reached the end of a line. > > littlebat > > http://www.learndiary.com/en/ > > > corvid wrote: > >Would it make things better or worse than it is now if I inserted a > >space when encountering a "?" or "?" so that dillo would know > >that it can wrap there? If that doesn't show up properly, > >I mean the characters that function as commas and periods. > > > >(I'm trying to think of something simple that will at least help you > > somewhat without having to make big changes right now :) > > I realize that it _should_ be able to wrap after most CJK characters, > > but...) > > > >It could be worse because > >1) Those spaces would still be there if you were copying and pasting > > the text to something else. > >2) If you wanted to search a page for "text?text", you would have to > > enter "text? text" with a space in there.

3 10

image memory leak
by Johannes.Hofmann＠gmx.de 13 May '09

13 May '09

Hi, I did some more tests regarding that image related memory leak, Jeremy reported some time ago. It turns out that memory related to aborted images is not released. With CSS and the related repushs this seems to happen frequently. But I can also trigger it without CSS by hitting Ctrl-R while loading a large image. Attached diagnostic patch shows the issue. Load a page with a single large JPEG image and hit Ctrl-R while it is loading. In that case only the second instance of DilloJpeg is deleted. I tried to figure out, why the first cache client is not getting aborted properly but the cache related code is quite complex... Does anyone know how this should work? Cheers, Johannes

3 8

towards a dillo-2.1 release
by jcid＠dillo.org 28 Apr '09

28 Apr '09

Hi there, It'd be very good if we can complete these points for the 2.1 release. Listed in no particular order: 1.- Dw documentation: online and as tarball. We have good docs. but to generate them there's a need to have: doxygen, graphviz, dot, latex and friends. That's why having it online and as a tarball is a good idea. 2.- Mapcity segfault In www.mapcity.cl playing with the "Tipo Calle" while the page is loading makes dillo segfault. If "Remote CSS" is disabled, it doesn't happen (also OK in dillo-2.0). 3.- New web site design We have a new banner and after finding a good color palette it may be ready to renew the site before 2.1 is announced. 4.- Debian/Ubuntu packages (after 2.1 final is ready) 5.- Update "8 years" page (it's nine) Yes, I should have done it already... 6.- Configurable key bindings. I'm reviewing this with Tim. Hopefully it will make it into the 2.1 release. 7.- Memory usage statistics Scott has posted some data on this, in another thread. I'll answer there. I'd be working on 5, 6 and 3. Volunteers please assign yourselves to the pending points. BTW, if you have important points to add to the list, please say so. Of course any segfaults or regressions have high priority too. TIA. -- Cheers Jorge.-

4 14

Image callback race condition
by onepoint＠starurchin.org 28 Apr '09

28 Apr '09

I am logging the creation and destruction of DilloImage structures (patches attached) and I have confirmed that sometimes the image callbacks run after the Image is freed. Here's a typical example: $ grep '^Image: 0x4687db8: ' ../../Log/log_2009-04-25_21-48-48.txt Image: 0x4687db8: new Image: 0x4687db8: ref: a_Jpeg_callback Image: 0x4687db8: free Image: 0x4687db8: ref: a_Jpeg_callback Image: 0x4687db8: ref: a_Jpeg_callback So how can I track down this race condition? Regards, Jeremy Henty PS: I am still running corvid's img-ref patch (also attached for reference).

2 2

Dillo documentation online
by onepoint＠starurchin.org 25 Apr '09

25 Apr '09

There's a tarball of the Dillo documentation at http://starurchin.org/dillo/dillo-docs.tar.gz Alternatively, download the link called "Documentation" at http://starurchin.org/dillo/ It will be updated every night, although it will only be as current as my local HG repo (which I update most days). Regards, Jeremy Henty

1 0

patch: a_Capi_get_flags and following redirection
by corvid＠lavabit.com 23 Apr '09

23 Apr '09

When an image on a page wouldn't load due to redirection, I would isolate the image, and often notice that message about Content-Length not matching message body. The reason is in a_Capi_open_url(), where reload is set by checking a_Capi_get_flags() for CAPI_IsCached. It follows the redirection, determines the Location URL isn't cached, and makes another HTTP request for the original URL. The cache entry data for that URL ends up containing the body from the first request with the header and body from the second request appended to it. Ah, but sometimes we want to follow redirection for get_flags. - the download check in capi - the load_stylesheet check in html.cc, assuming that we want to make that work with redirection at some point. - load_now for images in html.cc and the rest don't really matter too terribly much either way. So, unable to think of anything nice or clever, I added an argument to get_flags specifying whether redirection should be followed.

1 0

Instant client-side redirects (META refresh with delay=0)
by jcid＠dillo.org 18 Apr '09

18 Apr '09

Hi there, Please excuse me for being silent. Life has poured a few other important things to do over me (e.g. next week I have to move to another apartment). A patch for Instant client-side redirects is fresh in the repo. It took some time to develop because the whole idle/timeout mechanism needs a new API that allows to cancel queued functions. I didn't change the old API (the next release will keep it) but took some time to analyze and it may help with the redraw issue Johannes worked with, and also with hard-to-reproduce crashes. IIRC, Tim asked why several functions trigger from timeouts. Dillo is async. and the main driving function for I/O is Cache_process_queue(). This function is not reentrant, thus timeouts/idle calls allow for it to finish before a new action is taken. In general, there're several parts of dillo that need a more formal working definition or API. They've come to be with the implementation and it'd be good to define and document them, and hence not to have to go through all the involved code to find out how to implement something. I'll review the custom-keys patch series next, and try to re-start the work for the 2.1 release. Please test the patch. I've found it works well with the links in: http://git.kernel.org/gitweb.cgi -- Cheers Jorge.-

1 0