- Dillo-dev - lists.mailman3.com

Re: [Dillo-dev] wanted: CPU overhead source.

by Jorge Arellano Cid

On Fri, 5 Dec 2003, Johannes Hofmann wrote: > Hi Jorge, > > On Thu, Dec 04, 2003 at 10:43:34AM -0300, Jorge Arellano Cid wrote: > > > > I could not yet get any reproducible results from profiling, > > > > !? > > > > Doesn't it show any function usage-ratio/run-time change? > > > > Ok, I redid the profiling on my slower Linux box with utilities.html. > > Html_cleanup_tag is called more often in cvs version (16742 vs. 10749) and > seems quite a bit slower than in 0.7.3. > > See gprof output below. > > Hope that helps, > Johannes > > PS: Of course I can send you the whole gprof output, but it is quite > large, so I did not want to send the whole thing to the list. > [...] Thanks a lot! Now it's clear where's the problem. I'll review it together with another parsing issue in a few days I hope. Cheers Jorge.-

Dec. 11, 2003

latest network patch

by Jorge Arellano Cid

Hi there! A network patch that covers several aspects was just commited: + Connected the dpi stream to the cache using the CCC! + Fixed the cache API by introducing the new call a_Capi_get_buf(). + Fixed a race condition and a multiple request problem. + Enabled the https dpi (it doesn't work yet, just a development version). + Removed a cpu-hog when the network is unreachable. Besides what's stated above, it has improved a bit the network experience (for instance on sites with lots of small images. For ex. newspapers). Please test it and tell me how it works. Anders: Please try it and tell us if it also produces the locks you've seen. Does 0.7.3 also lock? Ferdi: It has the proper bindings between the cache and dpi, so HTML served over FTP works, among others! Some https sites are also viewable... Madis: This patch has the framework to implement https in a similar way than the ssl gateway. The connection between dillo and the https dpi is done by the generic dpi handler. Note that although it may be easy to add the POST method using wget to the current https dpi (wget >= 1.9.x using --post-data and the DilloUrl), making a native https plugin (avoiding wget) as in the ssl gatway is preferred. This allows for a simple way to implement connection caching. The current https dpi is a filter, not a server as it should to implement it this way. This is also easy to do. Regards Jorge.-

Dec. 11, 2003

Re: [Dillo-dev] Re: BUG#494

by Jorge Arellano Cid

On Wed, 10 Dec 2003, Anders Gavare wrote: > On Tue, 9 Dec 2003, Jorge Arellano Cid wrote: > > > Hi, > > > > Some time ago I read this bug and now the CVS has a bunch of > > fixed for signed/unsigned conflicts that may cause trouble on > > 64bit machines. Please test our CVS and tell me if the bug > > persists. > > Once I found out what the problem was, I mailed Sebastian Geerken about > it. I think that was in September. > > http://cvs.auriga.wearlab.de/cgi-bin/cvsweb.cgi/dillo/dillo/src/dw_widget.c… > > That fixed the rendering bug. OK. > I'm sorry that I forgot to remove the > entry from the bug database. I've not had any other problems on sparc64 > yet, except possible some random lock-ups. The version in CVS as of > 07:00 this morning seems to work just fine (that is, it still has the > random lock-ups but the output looks nice). lock-ups... Can you decribe them some more, or better, run dillo under gdb to find out where or why it locks (I haven't seen that bug ever). > I use dillo daily. It is very neat, especially on non-i386 platforms. > Keep up the good work :-) Thanks. Cheers Jorge.- PS: Are you on dillo-dev?

Dec. 10, 2003

BUG#494

by Andrew A. Gill

Good Ol' Jorge wrote: > > Some time ago I read this bug and now the CVS has a bunch of > fixed for signed/unsigned conflicts that may cause trouble on > 64bit machines. Please test our CVS and tell me if the bug > persists. Alpha EV56 running gcc 3.3.2 (glibc 2.3.2) reports no errors yet, sir! In fact, Slashdot now seems to be readable, thanks to something in Dillo CVS (with old Dillo it's still unreadable). -- |Andrew A. Gill |I posted to Silent-Tristero and| |<superluser(a)frontiernet.net> |all I got was this stupid sig! | |alt.tv.simpsons CBG-FAQ author | | | (Report all obscene mail to Le Maitre Pots)| |Nothing here yet: <http://www.needsfoodbadly.com> Temporary sig: -- Full speed ahead!

Dec. 10, 2003

Re: BUG#494

by Anders Gavare

On Tue, 9 Dec 2003, Jorge Arellano Cid wrote: > Hi, > > Some time ago I read this bug and now the CVS has a bunch of > fixed for signed/unsigned conflicts that may cause trouble on > 64bit machines. Please test our CVS and tell me if the bug > persists. Once I found out what the problem was, I mailed Sebastian Geerken about it. I think that was in September. http://cvs.auriga.wearlab.de/cgi-bin/cvsweb.cgi/dillo/dillo/src/dw_widget.c… That fixed the rendering bug. I'm sorry that I forgot to remove the entry from the bug database. I've not had any other problems on sparc64 yet, except possible some random lock-ups. The version in CVS as of 07:00 this morning seems to work just fine (that is, it still has the random lock-ups but the output looks nice). > Cheers > Jorge.- I use dillo daily. It is very neat, especially on non-i386 platforms. Keep up the good work :-) regards Anders

Dec. 10, 2003

navigation

by montegue＠myrealbox.com

i would like to see in addition to the forward and backward navigation buttons, one that goes to the parent directory, like konqueror does, and galeon if i am not mistaken :) thanks for a terrific application -- GR I have seen the future and it is just like the present, only longer. -- Kehlog Albran

Dec. 9, 2003

BUG#494

by Jorge Arellano Cid

Hi, Some time ago I read this bug and now the CVS has a bunch of fixed for signed/unsigned conflicts that may cause trouble on 64bit machines. Please test our CVS and tell me if the bug persists. You can send me the screenshot (not to dillo-dev please) Cheers Jorge.-

Dec. 9, 2003

Re: dpid

by Jorge Arellano Cid

On Sat, 6 Dec 2003, Ferdi Franceschini wrote: > Hi Jorge, > > On Fri, Dec 05, 2003 at 09:25:20AM -0300, Jorge Arellano Cid wrote: > > > > Hi Ferdi, > > > > I've catched an API problem inside the capi/cache that may be > > the cause for the fork-hog. This note is just to keep you from > > wasting time reviewing it. I'll write to you soon. > Great. It _was_ the cause! There was a race condition hidden behind the lack of dpi-cache connection (the part we blamed) and a slightly broken API (the cloaked problem). As a matter of fact this bug had remained unnoticed maybe for years. In my patched tree (which is under test), the bug is fixed (both parts) and dillo works even better than before. > I think I have a new solution to the inf loop caused when a connection is not > accepted on a dpi socket for any reason. This new solution does not require > any code changes to the plugins. I will post more info to dillo-dev but I need > to experiment a bit first. That's good news. Cheers Jorge.-

Dec. 9, 2003

Re: [Dillo-dev] wanted: CPU overhead source.

by Frank de Lange

BTW, The new kbnav patch is NOT on the site yet, wait for the announcement before you try to download it... Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Dec. 5, 2003

Re: [Dillo-dev] wanted: CPU overhead source.

by Frank de Lange

Related to this, I have a new version (probably the last, announcement to follow) of the kbnav patch which solves some performance issues related to g_list_search_custom(). This function, which caused an enormous overhead, has been eliminated from the patch alltogether in favor of a hashtable-based search. Frank -- WWWWW ________________________ ## o o\ / Frank de Lange \ }# \| / +46-734352015 \ \ `--| _/ <Hacker for Hire> \ `---' \ +31-640037120 / \ frank(a)unternet.org / `------------------------' [ "Omnis enim res, quae dando non deficit, dum habetur et non datur, nondum habetur, quomodo habenda est." ]

Dec. 5, 2003