October 2024 - Dillo-dev - lists.mailman3.com

Re: Rendering issue with view source and large embedded images

by Rodrigo Arias

Hi, On Sun, Oct 13, 2024 at 09:23:33PM +0200, a1ex(a)dismail.de wrote: >Rodrigo Arias <rodarima(a)gmail.com> wrote: > >> One thing that may be easy to try in the meanwhile is to run the same >> test by from a local HTTP server, so we don't use the file: DPI. If >> the bug persists, then it is not related with the file DPI. You can >> use a lightweight web server like darkhttpd. > >When opening the test page with the file dpi, the issue is very easy to >trigger. > >When using a local webserver as you suggested, it doesn't seem to >happen... > >However, for example when refreshing the site: >https://www.da.vidbuchanan.co.uk/blog/dram-emfi.html >I still see stuff like this: > >** WARNING **: CCC: call on already finished chain. Flags=CCC_Aborted >IO_write, closing with pending data not sent: >"ntZ1uGVvMRmC5GANpBI4569aV... > >And eventual glitching and crash after many refreshes. > >So, I'm not sure this is actually a dpi issue, but maybe that just >exposes it more. Ah, interesting. It seems to be time sensitive then. That's probably why I'm not able to see it. Do you have a particularly fast or slow machine? >To sum it up, I don't know shit, but guessing: >A double free is leading to a buffer overrun on the chain to the data >uri parser. Yes, there is a double free, but the question is: what leads to that condition? When Dillo detects this "CCC: call on already finished chain", it should abort, as this situation should never happen. Best, Rodrigo.

Oct. 14, 2024

Re: Rendering issue with view source and large embedded images

by a1ex＠dismail.de

Rodrigo Arias <rodarima(a)gmail.com> wrote: > One thing that may be easy to try in the meanwhile is to run the same > test by from a local HTTP server, so we don't use the file: DPI. If > the bug persists, then it is not related with the file DPI. You can > use a lightweight web server like darkhttpd. When opening the test page with the file dpi, the issue is very easy to trigger. When using a local webserver as you suggested, it doesn't seem to happen... However, for example when refreshing the site: https://www.da.vidbuchanan.co.uk/blog/dram-emfi.html I still see stuff like this: ** WARNING **: CCC: call on already finished chain. Flags=CCC_Aborted IO_write, closing with pending data not sent: "ntZ1uGVvMRmC5GANpBI4569aV... And eventual glitching and crash after many refreshes. So, I'm not sure this is actually a dpi issue, but maybe that just exposes it more. To sum it up, I don't know shit, but guessing: A double free is leading to a buffer overrun on the chain to the data uri parser. -Alex

Oct. 13, 2024

Re: Rendering issue with view source and large embedded images

by Rodrigo Arias

Hi, >I built and installed a clean Dillo with your file dpi patch. I think I will need to reproduce this to examine the logs and inspect the state of the CCC and the DPI while the problem occurs. I will try getting a OpenBSD installation at some point and debug this (and other problems with FreeBSD too). One thing that may be easy to try in the meanwhile is to run the same test by from a local HTTP server, so we don't use the file: DPI. If the bug persists, then it is not related with the file DPI. You can use a lightweight web server like darkhttpd. >Thats interesting, in my previous testing a few days ago I noticed >some messages about 'a_Tls_connection' in the debugger, which made no >sense since I'm working with a local file: Yes, a_Tls_connection() is always called and returns NULL when is not a TLS connection. Weird logic... Rodrigo.

Oct. 13, 2024

Re: Scrollbar themeing

by Rodrigo Arias

Hi, On Sun, Oct 13, 2024 at 05:56:26PM +0200, a1ex(a)dismail.de wrote: >Hi, > >When using a dark theme, the scrollbar can be very hard to see: > >https://alex.envs.net/test/scroll-theme-stock1.png > >I made some changes and now the scrollbar looks nicer and is more >visible: > >https://alex.envs.net/test/scroll-theme-mod1.png > >If someone wants to try it, here is a patch which also includes some >commented lines which you can use to further customize the appearance >of the scrollbar. I think this is a side effect of altering the colormap only partially. I would imagine that by mapping *all* light colors to dark counterparts, you will get coherent box borders. Here is more info: https://www.fltk.org/doc-1.3/drawing.html https://www.fltk.org/doc-1.3/drawing.html#drawing_colors This is an example of such colormap that may be easy to try using Fl:set_color(): https://github.com/fltk-rs/fltk-theme/blob/main/src/color_themes/dark.rs https://www.fltk.org/doc-1.3/group__fl__attributes.html#gac62f634ba8089a9d0… Best, Rodrigo.

Oct. 13, 2024

Re: Navigation by pages

by Rodrigo Arias

>Nice, this will probably become my default setting now. > >Thanks for this awesome patch! Thanks! Here are a couple of improvements: https://github.com/dillo-browser/dillo/pull/278 Clicking on the scrollbar arrows jumps to the top or bottom, and holding the button pressed will repeat the page scrolling. This only happens when the page scrolling mode is enabled (x)or Shift is hold. Please check the repeat and delay speeds, as I would rather prefer not to add more options to dillorc today :-) Also, do you usually set focus_new_tab to NO?, I'm considering switching the default to NO. Same with show_quit_dialog. Best, Rodrigo.

Oct. 13, 2024

Re: scroll_step description

by Rodrigo Arias

Hi, On Sat, Apr 20, 2024 at 01:54:26PM +1000, Kevin Koster wrote: >There's a new "scroll_step" option in the example configuration file for >3.1.0-rc1. The comment "Mouse wheel scroll step in pixels" suggests that >this only affects the mouse wheel, but it also changes the distance the >page moves when the arrow keys are pressed and the overlap of >previous page content that's left when you press Page Up/Down. I think >that's a little confusing, especially as the value has changed since >the default behaviour in 3.0.5 (the old value seems to have been about >25). This is now controlled independently with the scroll_page_overlap, which by default uses the value 50 px. While the scroll_step controls how much it advances with the arrows, by default 100 pixels. See: https://github.com/dillo-browser/dillo/pull/277 Best, Rodrigo.

Oct. 13, 2024

Scrollbar themeing

by a1ex＠dismail.de

Hi, When using a dark theme, the scrollbar can be very hard to see: https://alex.envs.net/test/scroll-theme-stock1.png I made some changes and now the scrollbar looks nicer and is more visible: https://alex.envs.net/test/scroll-theme-mod1.png If someone wants to try it, here is a patch which also includes some commented lines which you can use to further customize the appearance of the scrollbar. -Alex --- a/dw/fltkviewport.cc Sat Oct 12 22:05:30 2024 +++ b/dw/fltkviewport.cc Sun Oct 13 16:23:06 2024 @@ -59,12 +59,20 @@ FltkViewport::FltkViewport (int X, int Y, int W, int H hscrollbar = new CustScrollbar (x (), y (), 1, 1); hscrollbar->type(FL_HORIZONTAL); + // Horizontal scrollbar theme options: + // hscrollbar->color(FL_BLUE, FL_YELLOW); + // hscrollbar->box(FL_ENGRAVED_BOX); + hscrollbar->slider(FL_PLASTIC_DOWN_BOX); hscrollbar->callback (hscrollbarCallback, this); hscrollbar->hide(); add (hscrollbar); vscrollbar = new Fl_Scrollbar (x (), y(), 1, 1); vscrollbar->type(FL_VERTICAL); + // Vertical scrollbar theme options: + // vscrollbar->color(FL_BLUE, FL_YELLOW); + // vscrollbar->box(FL_ENGRAVED_BOX); + vscrollbar->slider(FL_PLASTIC_DOWN_BOX); vscrollbar->callback (vscrollbarCallback, this); vscrollbar->hide(); add (vscrollbar);

Oct. 13, 2024

Re: Navigation by pages

by a1ex＠dismail.de

Hi, > First, you can now scroll the mouse wheel over the vertical scrollbar > to jump whole pages instead of small steps. Over the page it > continues to behave as normal, unless you hold Shift, which will also > make it navigate full pages. This works really great, thanks! > Secondly, I also added a scrollbar "page mode", which can be enabled > in dillorc with scrollbar_page_mode=YES, which causes clicking with > the left button *anywhere* on the vertical scrollbar (including the > thumb) to scroll one page down. With the right button, one up. > Holding Shift can also enable temporarily this mode (or disable it). Another good feature, very useful if using a laptop or mouse without a proper scrollwheel. > I also added an option to place the vertical scrollbar on the left, > which I'm finding more convenient as it is usually closer to the text. Nice, this will probably become my default setting now. Thanks for this awesome patch! -Alex

Oct. 12, 2024

Re: Navigation by pages

by Rodrigo Arias

Hi, >Now having tried both, I feel that the button is still a bit nicer from >a usability standpoint, because it requires less attention and mouse >movement to go up and down. But, I agree there is not much point in >duplicating this functionality. I added a couple of options to improve the page navigation. First, you can now scroll the mouse wheel over the vertical scrollbar to jump whole pages instead of small steps. Over the page it continues to behave as normal, unless you hold Shift, which will also make it navigate full pages. Secondly, I also added a scrollbar "page mode", which can be enabled in dillorc with scrollbar_page_mode=YES, which causes clicking with the left button *anywhere* on the vertical scrollbar (including the thumb) to scroll one page down. With the right button, one up. Holding Shift can also enable temporarily this mode (or disable it). I also added an option to place the vertical scrollbar on the left, which I'm finding more convenient as it is usually closer to the text. Here is the PR that includes all the above, not merged yet: https://github.com/dillo-browser/dillo/pull/274 Best, Rodrigo.

Oct. 12, 2024

Re: Rendering issue with view source and large embedded images

by a1ex＠dismail.de

Hi Rodrigo, Rodrigo Arias <rodarima(a)gmail.com> wrote: > Place a breakpoint here: > > https://github.com/dillo-browser/dillo/blob/master/src/IO/IO.c#L189 > > Likely with "b IO.c:189", then "c" and "bt". > > That backtrace along with the stderr lines: > > >a_Dpi_ccc : OpSend [2F] Info=0xebed4db4980 Flags=0 > >a_Capi_ccc : OpSend [2F] Info=0xebee90cbd00 Flags=0 > > Can tell what is the state of the CCC at the moment that error > happens. > > The backtrace you attach happens later, so we already miss the point > where the problem occurs. I built and installed a clean Dillo with your file dpi patch. Here is that backtrace, which is from right before this: "WRITE Failed with -1: Broken pipe" [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 a_IO_ccc : OpSend [2F] Info=0xb5c0c2eba40 Flags=0 a_Dpi_ccc : OpSend [2F] Info=0xb5c0c30d000 Flags=0 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 a_Capi_ccc : OpSend [2F] Info=0xb5c0c30de00 Flags=0 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48be80, flags=5 state=12 a_IO_ccc : OpSend [2F] Info=0xb5c0c2eba40 Flags=0 a_Dpi_ccc : OpSend [2F] Info=0xb5c0c30d000 Flags=0 a_Capi_ccc : OpSend [2F] Info=0xb5c0c30de00 Flags=0 a_IO_ccc : OpEnd [2F] Info=0xb5c0c2eba40 Flags=0 a_Dpi_ccc : OpEnd [2F] Info=0xb5c0c30d000 Flags=0 a_Capi_ccc : OpEnd [2F] Info=0xb5c0c30de00 Flags=0 HTTP warning: Content-Length (937694) does NOT match message body (921310) for file:/tmp/dillo-doublefree.html a_Capi_ccc : OpEnd [1B] Info=0xb5c0c30d3c0 Flags=0 a_Dpi_ccc : OpEnd [1B] Info=0xb5c0c30d0c0 Flags=0 a_IO_ccc : OpEnd [1B] Info=0xb5c4883b140 Flags=0 Program received signal SIGPIPE, Broken pipe. _thread_sys_write () at /tmp/-:2 2 /tmp/-: No such file or directory. (gdb) bt #0 _thread_sys_write () at /tmp/-:2 #1 0x0ea252bb530f97da in ?? () #2 0x00000b5c41b50c52 in _libc_write_cancel (fd=7, buf=0xb5be6ed0000, nbytes=72966) at /usr/src/lib/libc/sys/w_write.c:27 #3 0x00000b597324b065 in IO_write (io=0xb5c48870ba0) at IO.c:230 #4 IO_callback (io=0xb5c48870ba0) at IO.c:275 #5 0x00000b597324adc9 in IO_fd_write_cb (fd=7, data=<optimized out>) at IO.c:317 #6 0x00000b5c1e858e88 in fl_wait(double) () from /usr/local/lib/libfltk.so.8.0 #7 0x00000b5c1e7e8a6e in Fl::wait(double) () from /usr/local/lib/libfltk.so.8.0 #8 0x00000b5c1e7e8c4d in Fl::run() () from /usr/local/lib/libfltk.so.8.0 #9 0x00000b59731f23b5 in main () (gdb) c Continuing. WRITE Failed with -1: Broken pipe a_IO_ccc : OpAbort [1F] Info=0xb5c4886d440 Flags=0 a_Dpi_ccc : OpAbort [1F] Info=0xb5c4886d180 Flags=0 a_Capi_ccc : OpAbort [1F] Info=0xb5c4883ba00 Flags=0 Here is another: a_Capi_ccc : OpStart [2B] Info=0xb5c0c2eba80 Flags=0 a_Dpi_ccc : OpStart [2B] Info=0xb5c48368640 Flags=0 a_IO_ccc : OpStart [2B] Info=0xb5c4886d400 Flags=0 a_Capi_ccc : OpStart [1B] Info=0xb5c4883b0c0 Flags=0 a_Dpi_ccc : OpStart [1B] Info=0xb5bc9415100 Flags=0 a_IO_ccc : OpStart [1B] Info=0xb5c483686c0 Flags=0 a_IO_ccc : OpSend [1B] Info=0xb5c483686c0 Flags=0 a_Capi_ccc : OpSend [1F] Info=0xb5c4883b0c0 Flags=0 a_Capi_ccc : OpSend [2B] Info=0xb5c0c2eba80 Flags=0 a_Dpi_ccc : OpSend [2B] Info=0xb5c48368640 Flags=0 a_IO_ccc : OpSend [2B] Info=0xb5c4886d400 Flags=0 a_Capi_ccc : OpSend [1F] Info=0xb5c4883b0c0 Flags=0 a_Capi_ccc : OpSend [1B] Info=0xb5c4883b0c0 Flags=0 a_Dpi_ccc : OpSend [1B] Info=0xb5bc9415100 Flags=0 a_IO_ccc : OpSend [1B] Info=0xb5c483686c0 Flags=0 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 a_IO_ccc : OpSend [2F] Info=0xb5c4886d940 Flags=0 a_Dpi_ccc : OpSend [2F] Info=0xb5c4886d9c0 Flags=0 a_Capi_ccc : OpSend [2F] Info=0xb5c0c30d980 Flags=0 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 a_IO_ccc : OpSend [2F] Info=0xb5c4886d940 Flags=0 a_Dpi_ccc : OpSend [2F] Info=0xb5c4886d9c0 Flags=0 a_Capi_ccc : OpSend [2F] Info=0xb5c0c30d980 Flags=0 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 Breakpoint 1, IO_read (io=0xb5c08fc3640) at IO.c:189 189 MSG("READ Failed with %d: %s\n", (int)St, strerror(errno)); (gdb) [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 bt #0 IO_read (io=0xb5c08fc3640) at IO.c:189 #1 IO_callback (io=0xb5c08fc3640) at IO.c:273 #2 0x00000b597324acc9 in IO_fd_read_cb (fd=7, data=0x51) at IO.c:294 #3 0x00000b5c1e858e88 in fl_wait(double) () from /usr/local/lib/libfltk.so.8.0 #4 0x00000b5c1e7e8a6e in Fl::wait(double) () from /usr/local/lib/libfltk.so.8.0 #5 0x00000b5c1e7e8c4d in Fl::run() () from /usr/local/lib/libfltk.so.8.0 #6 0x00000b59731f23b5 in main () (gdb) [file dpi]: File_serve_client 0x4b2ed48eb80, flags=5 state=12 (gdb) c Continuing. READ Failed with -1: Connection reset by peer a_IO_ccc : OpAbort [2F] Info=0xb5c4886d180 Flags=0 a_Dpi_ccc : OpAbort [2F] Info=0xb5c0c2eb880 Flags=0 ** WARNING **: Unused CCC Program received signal SIGPIPE, Broken pipe. _thread_sys_write () at /tmp/-:2 2 /tmp/-: No such file or directory. > This line: > > > READ Failed with -1: Connection reset by peer > > Is complaining that either a read() call or a Tls_read() failed. I > will assume it is the read() as there shouldn't be any https traffic > in your local test. It is communicating with the file DPI via a > socket. Thats interesting, in my previous testing a few days ago I noticed some messages about 'a_Tls_connection' in the debugger, which made no sense since I'm working with a local file: (gdb) break IO_write Breakpoint 1 at 0x9c0c8fc2fff: file IO.c, line 223. (gdb) c Continuing. WRITE Failed with -1: Broken pipe a_IO_ccc : OpAbort [1F] Info=0x9c335b28580 Flags=0 a_Dpi_ccc : OpAbort [1F] Info=0x9c335b00c40 Flags=0 a_Capi_ccc : OpAbort [1F] Info=0x9c335b00c00 Flags=0 Breakpoint 1, IO_write (io=0x9c38396eb40) at IO.c:223 223 void *conn = a_Tls_connection(io->FD); Regards, Alex

Oct. 12, 2024