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

Nav_repush() is hurting valgrind

by onepoint＠starurchin.org

On Sat, May 09, 2009 at 10:00:55AM -0400, Jorge Arellano Cid wrote: > On Sat, May 09, 2009 at 10:27:34AM +0100, Jeremy Henty wrote: > > > > The valgrind logs[1] show a recent burst of errors from calls to > > Nav_repush(). > > Oh, here I can't reproduce that. I've tried with some sites and > the valgrind log here is clean (a bit better than it used to be). > Do you have an example URL or test case? Not yet. I'll see what I can dig out of the logs. > BTW, I'm talking about the vanilla repository. > Are you testing with a patched repo? Yes. If it continues I'll back out any relevant patches and see if it changes anything. I've added a list of my local patches to the valgrind log page so that developers can see what I've done. http://starurchin.org/dillo/valgrind.html The only two that could (I think) possibly be relevant are: http://starurchin.org/dillo/patch/img_ref.diff.txt http://starurchin.org/dillo/patch/make-a_Cache_stop_client-close-client.txt The first patch was suggested to me by the list when I first reported the image access problems, the second is my attempt to fix it. I'd really appreciate any comments on whether they are doing things the right way. > The DNS here can't resolve the hostname. D'oh! My bad, that's the copy on my local machine. The correct links are above. Regards, Jeremy Henty

May 9, 2009

Nav_repush() is hurting valgrind

by jcid＠dillo.org

On Sat, May 09, 2009 at 10:27:34AM +0100, Jeremy Henty wrote: > > The valgrind logs[1] show a recent burst of errors from calls to > Nav_repush(). Oh, here I can't reproduce that. I've tried with some sites and the valgrind log here is clean (a bit better than it used to be). Do you have an example URL or test case? BTW, I'm talking about the vanilla repository. Are you testing with a patched repo? > Also, I'm currently trialling a patch to fix the image access errors > from a_Cache_stop_client() [2]. I don't think it can be the cause of > the Nav_repush() problems but I'd like a second opinion (I don't yet > fully understand how this code should work). > > [1] http://starurchin.singularity/dillo/valgrind.html > [2] http://starurchin.singularity/dillo/patch/make-a_Cache_stop_client-close-cl… The DNS here can't resolve the hostname. -- Cheers Jorge.-

May 9, 2009

[patch]: (updated) replace DW2TB with HT2TB

by jcid＠dillo.org

On Sat, May 09, 2009 at 10:52:32AM +0100, Jeremy Henty wrote: > > <headdesk> *Here's* the patch. Committed! -- Cheers Jorge.-

May 9, 2009

[patch]: (updated) replace DW2TB with HT2TB

by onepoint＠starurchin.org

<headdesk> *Here's* the patch. Jeremy Henty

May 9, 2009

[patch]: (updated) replace DW2TB with HT2TB

by onepoint＠starurchin.org

This is an updated version of the cleanup patch that replaces the convenience macro DW2TB with an even more convenient HT2TB. It doesn't change anything, but it reduces the noise in html.cc and form.cc . Please comment. Regards, Jeremy Henty

May 9, 2009

Nav_repush() is hurting valgrind

by onepoint＠starurchin.org

The valgrind logs[1] show a recent burst of errors from calls to Nav_repush(). Also, I'm currently trialling a patch to fix the image access errors from a_Cache_stop_client() [2]. I don't think it can be the cause of the Nav_repush() problems but I'd like a second opinion (I don't yet fully understand how this code should work). Regards, Jeremy Henty [1] http://starurchin.singularity/dillo/valgrind.html [2] http://starurchin.singularity/dillo/patch/make-a_Cache_stop_client-close-cl…

May 9, 2009

css and font specification [Re: Dillo2 + FLTK2 + NxLib running on mips linux]

by Johannes.Hofmann＠gmx.de

On Fri, May 08, 2009 at 07:59:43PM +0000, corvid wrote: > Johannes wrote: > > Hm, this seems to work fine for me. Maybe you have other stuff in > > your style.css that confuses the CSS parser? > > Yes, it turns out to have been an error in my style.css. > The previous rule was commented out, but I used html > comment delimiters instead of css comment delimiters. Sigh... > Sorry about that. All is well now. Not quite.... Html comment delimiters should be completely ignored by the CSS scanner/parser. With the re2c implementation that did work, but it is missing from the current scanner.

May 9, 2009

image memory leak

by jcid＠dillo.org

On Thu, Apr 23, 2009 at 06:01:56PM +0200, Hofmann Johannes wrote: > 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? The repository has a patch for this bug. It has the debug messages for easy testing. I've given it some stress but would like to see how it works in the outside world. Please give it some testing. -- Cheers Jorge.-

May 8, 2009

css and font specification [Re: Dillo2 + FLTK2 + NxLib running on mips linux]

by corvid＠lavabit.com

Johannes wrote: > Hm, this seems to work fine for me. Maybe you have other stuff in > your style.css that confuses the CSS parser? Yes, it turns out to have been an error in my style.css. The previous rule was commented out, but I used html comment delimiters instead of css comment delimiters. Sigh... Sorry about that. All is well now.

May 8, 2009

whitespace entities inside words

by corvid＠lavabit.com

Jorge wrote: > On Fri, May 08, 2009 at 03:53:03AM +0000, corvid wrote: > > Is a whitespace entity inside a word supposed to be > > treated like normal whitespace? > > > > I thought it would be, but first I saw that > > adjacent   ... aren't collapsed, > > then I realized that we are leaving the > > space inside the word instead of calling > > addSpace, and then I saw that PrevWasSPC > > isn't set if the space was at the end of the > > word. > > > > If it's just something that no one ever got > > around to dealing with, I'll make a patch. > > That hits into a deeper problem: the treatment of > whitespace is not well defined inside the textblock yet. > The same problem arises when handling NBSP. There's no > clear definition on how to handle/flag it. > If you're willing to study the textblock word wrapping, > and the whitespace problem, great! I think the next big > step for dillo is to handle DIV as a floating word (so it > can be rendered side by side, as intended), and AFAIS that > needs changes in the textblock. > This is a task for 2.2. I have some Textblock code around to fix "white-space: nowrap" (which shows up a lot on wikipedia, where the category stuff at the bottoms of pages is displayed badly) that I've been saving for discussion after this release. I have absolutely no clue about floating, though, and have been hoping that Johannes will develop an interest in it :)

May 8, 2009