dwb - a webkit web browser

karol · 2012-11-18 11:01:42

Lord Bo wrote:

Edit: Oh, I didn't know, that some topics can only be accessed if one is logged in. Well this goes for the post above.

It's because it's in off-topic section.

In my experience, dwb doesn't release memory after I close the websites that used it, it waits for the browser restart. Since dwb is very fast and light, restarting it once in a while isn't a terrible thing and takes just a few seconds.

Lord Bo · 2012-11-18 12:14:22

karol wrote:

In my experience, dwb doesn't release memory after I close the websites that used it, it waits for the browser restart. Since dwb is very fast and light, restarting it once in a while isn't a terrible thing and takes just a few seconds.

Well, I'm not complaining, I got a new laptop (well, rather new) and it can bear the memory leakage; Still, this is something one might want to fix sometime in the future.

Last edited by Lord Bo (2012-11-18 12:14:38)

WonderWoofy · 2012-11-18 15:50:59

Just wanted to let you know that I have been experiencing the browser crashes pretty consistently with dwb-hg.

These are back, though not as frequent:

** (dwb:12620): CRITICAL **: dwb_update_status_text: assertion `gl == dwb.state.fview' failed

Also, I figure I might as well just throw the ouput of the last crash in here as well.

 dwb

** (dwb:6507): CRITICAL **: dwb_update_status_text: assertion `gl == dwb.state.fview' failed
Received SIGSEGV, trying to clean up.

Last 24 stack frames: 

 24: dwb() [0x40cfa3]
 23: dwb() [0x41e03d]
 22: /usr/lib/libc.so.6(+0x35050) [0x7ff9ff032050]
 21: /usr/lib/libjavascriptcoregtk-1.0.so.0(_ZN3WTF10fastMallocEm+0xd4) [0x7ff9ffa318b4]
 20: /usr/lib/libjavascriptcoregtk-1.0.so.0(_ZN3WTF10StringImpl19createUninitializedEjRPh+0x2f) [0x7ff9ffa513bf]
 19: /usr/lib/libjavascriptcoregtk-1.0.so.0(_ZN3WTF10StringImpl6createEPKhj+0x38) [0x7ff9ffa555e8]
 18: /usr/lib/libjavascriptcoregtk-1.0.so.0(_ZN3WTF6StringC2EPKcj+0x15) [0x7ff9ffa56c45]
 17: /usr/lib/libwebkitgtk-1.0.so.0(+0xb19356) [0x7ffa01de1356]
 16: /usr/lib/libwebkitgtk-1.0.so.0(+0xb1ee99) [0x7ffa01de6e99]
 15: /usr/lib/libwebkitgtk-1.0.so.0(+0x3fda1c) [0x7ffa016c5a1c]
 14: /usr/lib/libwebkitgtk-1.0.so.0(+0xa22232) [0x7ffa01cea232]
 13: /usr/lib/libwebkitgtk-1.0.so.0(+0x9e9aa6) [0x7ffa01cb1aa6]
 12: /usr/lib/libwebkitgtk-1.0.so.0(+0x9c5472) [0x7ffa01c8d472]
 11: /usr/lib/libwebkitgtk-1.0.so.0(+0x9c6cc3) [0x7ffa01c8ecc3]
 10: /usr/lib/libwebkitgtk-1.0.so.0(+0x708c16) [0x7ffa019d0c16]
  9: /usr/lib/libwebkitgtk-1.0.so.0(+0x70bc5a) [0x7ffa019d3c5a]
  8: /usr/lib/libwebkitgtk-1.0.so.0(+0x70bfd9) [0x7ffa019d3fd9]
  7: /usr/lib/libwebkitgtk-1.0.so.0(+0x866255) [0x7ffa01b2e255]
  6: /usr/lib/libwebkitgtk-1.0.so.0(+0x695303) [0x7ffa0195d303]
  5: /usr/lib/libwebkitgtk-1.0.so.0(+0x695846) [0x7ffa0195d846]
  4: /usr/lib/libwebkitgtk-1.0.so.0(+0x6ef940) [0x7ffa019b7940]
  3: /usr/lib/libwebkitgtk-1.0.so.0(+0x5430ee) [0x7ffa0180b0ee]
  2: /usr/lib/libwebkitgtk-1.0.so.0(+0x115a13b) [0x7ffa0242213b]
  1: [0x7ff9b35eb265]
[1]    6507 segmentation fault (core dumped)  dwb
dwb  51.17s user 6.49s system 0% cpu 11:41:51.60 total

Of course, let me know if there is more information I might be able to collect for you. I love this browser and am happy to be a test subject to help you make it better. Unforuntaely, that is the extent to which I am able to help.

BTW, @Lord Bo, You linked to the other post, and I thought this issue might have been discussed in that thread. The only thing there was you saying the exact same thing. That was really of no help.

Last edited by WonderWoofy (2012-11-18 16:51:54)

oeufcoque · 2012-11-18 16:41:18

".config/dwb/default/session" sometimes gets garbage written to it. The sessions works still, it's just that there are additional garbage sessions written to it.

Is anyone else experiencing this?

With garbage I mean a session I haven't added, named something like "?+"

Here is the relevant line in the session file:

g:?+^A

Lord Bo · 2012-11-18 16:45:28

@WonderWoofy: Sorry, but what I posted was completely unrelated to the issue. And also your backtrace looks a bit strange. Maybe something with the backtrace script is not correct? I will have a look at it. However: This assertion failure seems to stem from dwb.c:599 g_return_if_fail(gl == dwb.state.fview); .

WonderWoofy · 2012-11-18 16:53:54

Lord Bo, are you looking at my backtrace from post #1274, or what I posted a couple posts up? Because a couple posts up is just what was dumped into my terminal when the last crash occured. I had been starting dwb from the terminal so that I might be able to get some useful information when it crashes.

webspider · 2012-11-18 16:58:53

I get similar crashes like reported in post #1226, they seem to happen preferably when either an image in a tab is still loading, it's a very big image or it's a very big gif. Even if I open another adress in the tab the image was in, closing it causes dwb to crash. I tried switching between dwb, dwb-hg, dwb-gtk3 and dwb-gtk3-hg, but this doesn't really help and seems to affect the gtk3 versions more. If there is anything I can do for debugging, tell me, because X just outputs generic messages like in post #1226.

edit: Output from backtrace.sh

Last edited by webspider (2012-11-18 17:05:53)

portix · 2012-11-18 17:01:01

WonderWoofy wrote:

Of course, let me know if there is more information I might be able to collect for you. I love this browser and am happy to be a test subject to help you make it better. Unforuntaely, that is the extent to which I am able to help.

As i already said a backtrace would be helpful, but only backtraces where dwb has crashed.

Lord Bo · 2012-11-18 17:02:41

WonderWoofy wrote:

Lord Bo, are you looking at my backtrace from post #1274, or what I posted a couple posts up? Because a couple posts up is just what was dumped into my terminal when the last crash occured. I had been starting dwb from the terminal so that I might be able to get some useful information when it crashes.

Ok, sorry again, I was a bit confused since it didn't look like what I had expected. I'm now trying to use the mercurial version to, hopefully the browser will crash, too.

Lord Bo · 2012-11-18 17:04:05

portix wrote:

WonderWoofy wrote:
Of course, let me know if there is more information I might be able to collect for you. I love this browser and am happy to be a test subject to help you make it better. Unforuntaely, that is the extent to which I am able to help.
As i already said a backtrace would be helpful, but only backtraces where dwb has crashed.

So it would be convenient to invoke dwb with gdb dwb and then wait until it crashes to invoke 'backtrace' from inside gdb? Just to sum it up.

Lord Bo · 2012-11-18 17:08:53

Ohhh... Your script is supposed to do that, but it invokes make debug and doing that fails! I just compiled dwb with a simple make and gdb does not complain... .
No, it is allright. Just didn't notice that it invoked make debug from within the src directory.

Edit: There is something wrong with the Makefile. It is not complete. There is no debug version, but the normal version is compiled with the -g flag, whilst -O2 optimization is given. That is a bit strange... I think you should do something about that. That then seems not to lead to anything, too.

Last edited by Lord Bo (2012-11-18 17:47:54)

WonderWoofy · 2012-11-18 17:22:33

Oh, okay, I get it. Sorry portix. I will try to do better now. Gotta reboot anyway after a kenel update from graysky. I will start it t via your script and let it run and try and get some data for you.

Lord Bo · 2012-11-18 18:28:57

Ok portix, I got one for you. It is easily to reproduce (at least for me): Open dwb:keys and press Return. The browser will crash. I appended the log. However, I think this has rather low priority, if it is not related to the other crashes. gdb.log

portix · 2012-11-18 18:48:19

Lord Bo wrote:

Ok portix, I got one for you. It is easily to reproduce (at least for me): Open dwb:keys and press Return. The browser will crash. I appended the log. However, I think this has rather low priority, if it is not related to the other crashes. gdb.log

Thanks, fixed.

Lord Bo · 2012-11-18 19:09:29

portix wrote:

Thanks, fixed.

Wow, that was fast, thank you.

WonderWoofy · 2012-11-19 02:58:01

Okay, I think I did better this time. I ran the script and used the browser it produced until it crashed. Here is the gdb.log.

Last edited by WonderWoofy (2012-11-19 02:58:21)

Lord Bo · 2012-11-19 10:15:36

Another Bug: dwb does not paste from primary if shift+insert is pressed.

portix · 2012-11-19 19:09:21

@WonderWoofy: Thanks, but that webkit bug.

@Lord Bo: That's not a bug, all gtk-applications don't paste on shift-insert. In dwb-hg i tried to implement shift insert but it doesn't always work because gtk sometimes returns the clipboard even if primary selection is requested.

webspider · 2012-11-19 20:12:43

@portix: What about this one regarding closing loading images, large images and large gifs? I hope that's not webkit specific.

portix · 2012-11-19 20:20:03

webspider wrote:

@portix: What about this one regarding closing loading images, large images and large gifs? I hope that's not webkit specific.

I'm not sure but it seems to be webkit related, dwb calls gtk_widget_destroy but a the webview must be destroyed when the tab is closed. Do you have accelerated-compositing enabled? enable-accelerated-compositing is buggy in webkitgtk and it causes dwb and X to crash sometimes.

Lord Bo · 2012-11-20 09:01:43

portix wrote:

@Lord Bo: That's not a bug, all gtk-applications don't paste on shift-insert. In dwb-hg i tried to implement shift insert but it doesn't always work because gtk sometimes returns the clipboard even if primary selection is requested.

That's funny!

Lord Bo · 2012-11-20 10:01:14

Ok, I'm trying to write some scripts, which shall be used with dwb. Doing that I got a little problem: I'd like to bind a keycombination to some function. This function has one or more parameters. Now the bind shall take that into account. Look at the following example:

#!javascript

function testprom(val) 
{
  var ret = io.prompt("testprom");
  if( ret != null )
  {
    io.error(val);
  }
}

bind("ex", testprom("No error occured!") );

Ok what I'd like to happen is, that if the user presses an 'e' followed by an 'x' and does not cancel at the prompt, the error message "No error occurred" appears. So how can I pass parameters together with the callback function to bind() ?

Last edited by Lord Bo (2012-11-20 10:01:53)

portix · 2012-11-20 10:19:12

Lord Bo wrote:

Ok, I'm trying to write some scripts, which shall be used with dwb. Doing that I got a little problem: I'd like to bind a keycombination to some function. This function has one or more parameters. Now the bind shall take that into account. Look at the following example:
#!javascript

function testprom(val) 
{
  var ret = io.prompt("testprom");
  if( ret != null )
  {
    io.error(val);
  }
}

bind("ex", testprom("No error occured!") );
Ok what I'd like to happen is, that if the user presses an 'e' followed by an 'x' and does not cancel at the prompt, the error message "No error occurred" appears. So how can I pass parameters together with the callback function to bind() ?

You have different possibilities, you can use a closure

bind("ex", function() { testprom("No error occured!"); });

or the native bind method that all functions have

bind("ex", testprom.bind(this, "No error occured!"));

or the curry method that dwb implements for every function

bind("ex", testprom.curry("No error occured");

The mistake in your code is, bind expects a function but if you just write testprom("No error occured"), testprom is executed immediately and the argument of bind is the return value of testprom which will be undefined in your example.

Edit: For function.curry you need dwb-hg.

Edit2: After writing this post i came to the conclusion that i should remove curry since bind does the same, so you should either use bind or the anonymous function.

Last edited by portix (2012-11-20 10:29:02)

Lord Bo · 2012-11-20 10:45:39

Thanks a lot for this good explanation!

webspider · 2012-11-20 20:46:47

@portix: I don't have it enabled, only a compositing manager. After I deactivated that, the resulting backtrace looks slightly different. Damn it.

Arch Linux

#1276 2012-11-18 11:01:42

Re: dwb - a webkit web browser

#1277 2012-11-18 12:14:22

Re: dwb - a webkit web browser

#1278 2012-11-18 15:50:59

Re: dwb - a webkit web browser

#1279 2012-11-18 16:41:18

Re: dwb - a webkit web browser

#1280 2012-11-18 16:45:28

Re: dwb - a webkit web browser

#1281 2012-11-18 16:53:54

Re: dwb - a webkit web browser

#1282 2012-11-18 16:58:53

Re: dwb - a webkit web browser

#1283 2012-11-18 17:01:01

Re: dwb - a webkit web browser

#1284 2012-11-18 17:02:41

Re: dwb - a webkit web browser

#1285 2012-11-18 17:04:05

Re: dwb - a webkit web browser

#1286 2012-11-18 17:08:53

Re: dwb - a webkit web browser

#1287 2012-11-18 17:22:33

Re: dwb - a webkit web browser

#1288 2012-11-18 18:28:57

Re: dwb - a webkit web browser

#1289 2012-11-18 18:48:19

Re: dwb - a webkit web browser

#1290 2012-11-18 19:09:29

Re: dwb - a webkit web browser

#1291 2012-11-19 02:58:01

Re: dwb - a webkit web browser

#1292 2012-11-19 10:15:36

Re: dwb - a webkit web browser

#1293 2012-11-19 19:09:21

Re: dwb - a webkit web browser

#1294 2012-11-19 20:12:43

Re: dwb - a webkit web browser

#1295 2012-11-19 20:20:03

Re: dwb - a webkit web browser

#1296 2012-11-20 09:01:43

Re: dwb - a webkit web browser

#1297 2012-11-20 10:01:14

Re: dwb - a webkit web browser

#1298 2012-11-20 10:19:12

Re: dwb - a webkit web browser

#1299 2012-11-20 10:45:39

Re: dwb - a webkit web browser

#1300 2012-11-20 20:46:47

Re: dwb - a webkit web browser

Board footer