- Dillo-dev - lists.mailman3.com

Delayed release

by Jorge Arellano Cid

Hi there, I just commited the code that cleanups the bookmarks server. It works like this: <q> /* * If this is our first run, kill the (old) bookmarks server, * and create the '~/.dillo/.072' file to flag what we did. */ static void Dillo_check_bm_srv12(char *dir) [...] </q> Yes, it uses the "[b]m_srv12" trick and some other precautions. Please test it: the whole point is that when installing the new version of dillo (072), the old bookmarks server may keep running and survive the updated. The new code detects the first dillo-072 run and issues a kill command to "bm_srv12". Onto another topic: A last minute leak fix (iterators) left dillo a bit unstable. We knew the iterators had a leak and it's great to have found it. Now it will take some extra time to try to fix the problem and, after that, stress-test the Copy&Paste code again. We haven't got an accurate procedure to reproduce the crash (yet). This explanation may help: <q> I think I have found a relatively new segfault bug : You need some sort of thumbtags like the screenshots on http://www.incagold.com/products/Speed/speedthief.php Now you have to move down the mouse and at the same time click on one image (i.e. click while still aiming for the image). If the button-push occures outside the image and the button-release inside the image, chances are high to get a segfault. This is a bit hard to reproduce, but I couldn't get an older version with selection to behave like that. It also happened a few times to me last night (after the last 'cvs up') but never before. </q> We'll be working on it. Cheers Jorge.-

April 24, 2003

Re: [Dillo-dev] Cookie locks?

by Gabriel Kihlman

Hi, Lars Clausen <lrclause(a)cs.uiuc.edu> writes: > I understand unrestricted access to cookie jars can be bad for your health, > but it's annoying that you can't have two Dillos using cookies at the same > time. I usually have a Dillo running permanently, and when I get a URL in > my mail, I can click on it to open another Dillo (quickly!) -- but that > Dillo can't use cookies, which is frequently (e.g. in bugzilla) necessary. > Could we do a write-lock on it? That is something I would appreciate as well. I have been running with this ugly patch that disables locking for atleast 6-7 months without a problem: --- src/cookies.c.orig Sun Jan 12 01:25:20 2003 +++ src/cookies.c Mon Feb 24 09:43:37 2003 @@ -31,7 +31,6 @@ #include "list.h" #include "cookies.h" -#define DEBUG_LEVEL 8 #include "debug.h" /* The maximum length of a line in the cookie file */ @@ -140,7 +139,7 @@ void a_Cookies_init() char line[LINE_MAXLEN]; /* Default setting */ - disabled = TRUE; + disabled = FALSE; /* Read and parse the cookie control file (cookiesrc) */ if (Cookie_control_init() != 0) { @@ -158,6 +157,7 @@ void a_Cookies_init() return; } +#if 0 /* Try to get a lock from the file descriptor */ disabled = (lockf(fileno(file_stream), F_TLOCK, 0) == -1); if (disabled) { @@ -166,7 +166,7 @@ void a_Cookies_init() fclose(file_stream); return; } - +#endif DEBUG_MSG(10, "Enabling cookies as from cookiesrc...\n"); cookies = g_hash_table_new(g_str_hash, g_str_equal);

April 24, 2003

Re: [Dillo-dev] Vera

by Nicola Girardi

> > Yes, I'm using it now. > > I can use it too, but it's not anti-aliased. xterm is able to use > anti-aliased TTF fonts (seemingly ONLY Luxi, but that's another issue), > so I know my X setup can use it. What do you have to do to get > anti-aliasing? Apply my patch, look in the archives. Should apply with minor tweaks. (Again, I hope this creeps in the official version.)

April 23, 2003

Re: [Dillo-dev] Vera works now

by Andreas Zwinkau

> Sounds like the "FontPath" settings in your XF86Config file do not > match the "catalogue" settings in your xfs config > (<X11etcdir>/fs/config). xfs is simply a decent cache, which can > provide fonts over the network(only need to install on one machine, > etc etc). Update your server config to use more font paths. You can > augment the list at run time with "xset fp+ /path/to/more/fonts" if > you don't want to touch XF86Config. > > One thing is to make sure that you run "xset fp rehash" after adding > new fonts and running "mkfontdir", since X doesn't automatically see > new fonts. I should have said that before, but only gave "xfstt > --sync". Sorry. I thought it was a xfs problem, so i did run it. But now i checked my X output and it says "Could not init font path element unix/:7100, removing from list!". So xfs is not used! I think the problem was mkfontdir. I thought fc-cache (gentoo) would do this, but it does not seem so. Xfs did mkfontdir or something similar on all directories and so Dillo can find them now. Thats my assumption. I checked .../fs/config and it is there. Thanks for help :D -- Andreas Zwinkau | web: andi.dasstellenwirinsinternet.de | mail: andi(a)buxach.de | jabber: beza1e1(a)amessage.de

April 23, 2003

Re: [Dillo-dev] Vera

by Phil Pennock

On 2003-04-22 at 18:25 +0200, Andreas Zwinkau wrote: > I did not use xfs, was that the problem? Mozilla and xchat support Vera > nevertheless. Now i run xfs Dillo and xlsfonts do find Vera, too. Sounds like the "FontPath" settings in your XF86Config file do not match the "catalogue" settings in your xfs config (<X11etcdir>/fs/config). xfs is simply a decent cache, which can provide fonts over the network (only need to install on one machine, etc etc). Update your server config to use more font paths. You can augment the list at run time with "xset fp+ /path/to/more/fonts" if you don't want to touch XF86Config. One thing is to make sure that you run "xset fp rehash" after adding new fonts and running "mkfontdir", since X doesn't automatically see new fonts. I should have said that before, but only gave "xfstt --sync". Sorry. So it could simply be fixed by having restarted X, since that effectively did a rehash. ;^) -Phil

April 23, 2003

Re: [Dillo-dev] Unix domain sockets

by Phil Pennock

On 2003-04-22 at 16:20 -0400, Jorge Arellano Cid wrote: > kill `ps -U <USER> | grep bm_srv12 | sed 's/^ *//g' | cut -d' ' -f1` Depending upon line buffering and how your PIDs are ordered (wrapped around recently?) that can end up killing the "grep" process before its given the PID for the bookmark server. A neat little Unix shell idiom for that is: grep '[b]m_srv12' The pattern there will no longer match the grep command-line, since the ']' adds an extra character. I think that I first learnt this from Aeleen Frisch's Essential System Adminstration book -- an excellent resource now in its third edition (was 1st ed when I learnt that one -- it's an old idiom). ps -U <USER> | grep '[b]m_srv' | awk '{print $1}' | xargs kill > That is, issued from dillo to force the shut down of a running > bookmark server. Suggestion: put it into a two-line shell-script which Dillo invokes; then portability simply becomes a package-maintainer issue of changing a very short shell script. > The point is: Does this command work well on *BSDs, and > Solaris? The basic stuff tested successfully on FreeBSD 4.7 ia32, OpenBSD 3.2 sparc and Solaris 2.5.1 sparc. Both the sed/cut version and the awk version. ;^) That's using the default /usr/bin/ps on Solaris; if someone has put the *BSD commands in their path first, then they'll have the really ancient syntax of /usr/ucb/ps and -U isn't supported. Such people are probably best just deemed "on their own".

April 23, 2003

Re: [Dillo-dev] Unix domain sockets

by Andreas Schweitzer

On Wed, Apr 23, 2003 at 11:35:16AM +0200, Lars Segerlund wrote: > > This is actually a newbie error on solaris :-) ... killall does kill > ALL ! same as HPUX, think AIX is the same. Yep, AIX is just as "destructive" :-) > >>> kill `ps -U <USER> | grep bm_srv12 | sed 's/^ *//g' | cut -d' ' -f1` Oh, and while being with AIX, this will NOT work on AIX. The first field on AIX is the UID ... But dillo doesn't compile easily on AIX anyways :-). Back when I was working on an AIX machine I did have an executable, but there were always some gotchas. Cheers, Andreas -- **************************** NEW ADDRESS ****************************** Hamburger Sternwarte Universitaet Hamburg Gojenbergsweg 112 Tel. ++49 40 42891 4016 D-21029 Hamburg, Germany Fax. ++49 40 42891 4198

April 23, 2003

Re: [Dillo-dev] Unix domain sockets

by Lars Segerlund

This is actually a newbie error on solaris :-) ... killall does kill ALL ! same as HPUX, think AIX is the same. / lars Segerlund Lars Clausen wrote: > On Tue, 22 Apr 2003, Andreas Schweitzer wrote: > >>Hi, >> >> >>> I kept cosidering the alternatives, and apart from the UDS idea >>>(not feasible), the incomplete "bye" command, there's also >>>another option! >>> >>> kill `ps -U <USER> | grep bm_srv12 | sed 's/^ *//g' | cut -d' ' -f1` >> >>On OpenBSD and FreeBSD, this construct does return the PID >>indeed. I'm just not sure how portable that is in general. >> >>Quite less portable is killall (which does not even exist on OpenBSD), >>but on FreeBSD it has an -u for user option. I.e. "killall -u andy >>bm_srv12" will do it, but only on FreeBSD :-( > > > Don't ever, *ever* consider killall portable. Under at least HPUX, killall > silently ignores its arguments and kills all the processes it can get its > murderous little hands on. A bad day to be root... > > -Lars >

April 23, 2003

Re: [Dillo-dev] Some thoughts

by Lars Segerlund

I agree, I have been trying to 'break out' the bookmark dpi into a loadable object file, ( loads at runtime, and installs the button and such ), in order to do work on a dpi plugin myself. If something about loading dpi modules were in place it would be much nicer. I have tried to hack a list of dpi modules in .dillo/dpis which is loaded at runtime, but so far not much sucess. Is there a plan on how these things will work in the future ? Because then one can progress with the dpi itself and the infrastructure. / Lars Segerlund. Andreas Schweitzer wrote: > Hi, > > Just a comment on one of your plans : > > >> Appart from that, I still have to finish the design of dpi >>initialization, addition, managing etc. (needed to support >>several dpis at the same time). > > > I think this will be important, since it kept me somewhat from > playing with dpi's. I did some experiments, but an automatic > dpi initilization makes coding much easier. And I have the feeling > others think the same :-) > > Something like > dpi:/foo/ > starts the dpi foo[possible-extra-name-part] from some location like > $PATH or some other "repository". And > dpi:/foo/command > passes "command" to the dpi foo. > Just some rambling :-) > > Cheers, > Andreas >

April 23, 2003

Re: [Dillo-dev] Vera

by Chris Palmer

Phil Pennock writes: > > Example: http://www.bitstream.com/categories/products/fonts/vera/ > > Is it possible to use it in Dillo? > > Yes, I'm using it now. I can use it too, but it's not anti-aliased. xterm is able to use anti-aliased TTF fonts (seemingly ONLY Luxi, but that's another issue), so I know my X setup can use it. What do you have to do to get anti-aliasing? -- Chris music is what numbers feel like San Francisco, CA http://www.nodewarrior.org/chris

April 23, 2003