- Dillo-dev - lists.mailman3.com

image memory leak

by onepoint＠starurchin.org

On Thu, Apr 23, 2009 at 06:01:56PM +0200, Hofmann Johannes wrote: > It turns out that memory related to aborted images is not released. What about the other way round, ie. memory being released and then accessed? If you check the invalid read errors at the valgrind logs page[1] for the past few days you will see they are all due to reading freed image memory. Is this another aspect of the same thing, or something else? > 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? No, I found that part of the code rather brainbending I'm afraid. Regards, Jeremy Henty [1] http://starurchin.org/dillo/valgrind.html

April 23, 2009

image memory leak

by Johannes.Hofmann＠gmx.de

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

April 23, 2009

patch: a_Capi_get_flags and following redirection

by corvid＠lavabit.com

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.

April 23, 2009

towards a dillo-2.1 release

by jcid＠dillo.org

Hi Devid, On Wed, Apr 22, 2009 at 03:05:37PM +0200, Devid Antonio Filoni wrote: > Hi, > I can work on point 4. Good! Although packages are to be provided after the final release is ready, you should start testing the packages now. Just put them somewhere (based on current repo) and let the list know for people to check they work OK. If you don't have a server to upload them, we can host them in dillo.org/test. > > Devid Antonio Filoni > > On Mon, Apr 20, 2009 at 11:10 PM, Jorge Arellano Cid <jcid(a)dillo.org> wrote: > > 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. -- Cheers Jorge.-

April 22, 2009

towards a dillo-2.1 release

by devidfil＠gmail.com

Hi, I can work on point 4. Devid Antonio Filoni On Mon, Apr 20, 2009 at 11:10 PM, Jorge Arellano Cid <jcid(a)dillo.org> wrote: > 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.- > > _______________________________________________ > Dillo-dev mailing list > Dillo-dev(a)dillo.org > http://lists.auriga.wearlab.de/cgi-bin/mailman/listinfo/dillo-dev >

April 22, 2009

towards a dillo-2.1 release

by Johannes.Hofmann＠gmx.de

On Tue, Apr 21, 2009 at 11:06:09AM -0400, Jorge Arellano Cid wrote: > On Tue, Apr 21, 2009 at 09:01:46AM +0200, Hofmann Johannes wrote: > > On Tue, Apr 21, 2009 at 07:48:11AM +0200, Hofmann Johannes wrote: > > > On Mon, Apr 20, 2009 at 05:10:41PM -0400, Jorge Arellano Cid wrote: > > > > 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. > > > > > > I will check 2. > > > > Ok, I can reproduce it now with > > > > trickle -d 15 dillo > > > > trickle (http://monkey.org/~marius/pages/?page=trickle) is really > > cool for debugging! > > > > The problem is the repush caused by the arriving CSS stylesheet. It > > deletes all form widgets and recreates them later. > > I guess I can fix the segfault somehow, but there is a fundamental > > problem if the user starts to fill in a form and then the repush > > deletes all his data :( > > > > I currently see these options: > > > > 1. Disable forms while CSS stylesheets are pending. > > 2. Disable the repush once the user started to fill in forms. > > 3. Save form data in memory and restore it after repush. > > This could also be used to restore form data when going back to a > > forms page. > > > > Opinions? > > To implement what we had with dillo1: This is, inactive forms > until the full form is loaded. Adding the restriction of pending > stylesheets too (like in 1). Ok, I just commited a fix. Please test carefully as there might be negative sideeffects like permanentely deactivated forms if a stylesheet fails to load. Cheers, Johannes

April 22, 2009

towards a dillo-2.1 release

by jcid＠dillo.org

On Tue, Apr 21, 2009 at 08:03:57PM +0000, corvid wrote: > Jorge wrote: > > Please check and comment. > > > The next release, dillo-2.1, scheduled for middle 2009 will provide basic CSS > > support, and the usual bugfixes and cleanups. > > Auth deserves mention, I think. > And I'd add "new features" or "improvements" in front of > "bugfixes and cleanups", since we have added various new > little things as well. Please go ahead. -- Cheers Jorge.-

April 21, 2009

towards a dillo-2.1 release

by corvid＠lavabit.com

Jorge wrote: > Please check and comment. > The next release, dillo-2.1, scheduled for middle 2009 will provide basic CSS > support, and the usual bugfixes and cleanups. Auth deserves mention, I think. And I'd add "new features" or "improvements" in front of "bugfixes and cleanups", since we have added various new little things as well.

April 21, 2009

towards a dillo-2.1 release

by jcid＠dillo.org

On Mon, Apr 20, 2009 at 05:10:41PM -0400, Jorge Arellano Cid wrote: > 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. 5 is done. BTW, there's a good potential link for "memory usage" when the collected data makes its way into a web page. Please check and comment. -- Cheers Jorge.-

April 21, 2009

towards a dillo-2.1 release

by corvid＠lavabit.com

Jorge wrote: > On Mon, Apr 20, 2009 at 10:46:49PM +0000, corvid wrote: > > Jorge wrote: > > > Of course any segfaults or regressions have high priority too. > > > > The dicache changes broke scaled multiscan images. > > I don't think FltkImgbuf::newScan() gets called anymore. > > You're right. Can you look at it? > > BTW: FltkImgbuf::newScan() can be issued by adding a call to > a_Image_new_scan(). All right. PS I was glad to hear that the website will get a new banner.

April 21, 2009