- Dillo-dev - lists.mailman3.com

hyphenator valgrind error

by corvid＠lavabit.com

==30983== Invalid read of size 1 ==30983== at 0x81151B6: fl_utf8back (fl_utf.c:259) ==30983== by 0x80A2320: dw::fltk::FltkPlatform::prevGlyph(char const*, int) (fltkplatform.cc:599) ==30983== by 0x809EDA3: dw::Hyphenator::hyphenateSingleWord(dw::core::Platform*, char*, int, lout::misc::SimpleVector<int>*) (hyphenator.cc:327) ==30983== by 0x809E9F2: dw::Hyphenator::hyphenateWord(dw::core::Platform*, char const*, int*) (hyphenator.cc:256) ==30983== by 0x809C087: dw::Textblock::hyphenateWord(int) (textblock_linebreaking.cc:762) ==30983== by 0x809BB6E: dw::Textblock::wordWrap(int, bool) (textblock_linebreaking.cc:609) ==30983== by 0x809B6A9: dw::Textblock::processWord(int) (textblock_linebreaking.cc:441) ==30983== by 0x8098753: dw::Textblock::addText0(char const*, unsigned int, short, dw::core::style::Style*, dw::core::Requisition*) (textblock.cc:1694) ==30983== by 0x8097B02: dw::Textblock::addText(char const*, unsigned int, dw::core::style::Style*) (textblock.cc:1478) ==30983== by 0x8073F5C: Html_process_word(DilloHtml*, char const*, int) (html.cc:1216) ==30983== by 0x807A348: Html_write_raw(DilloHtml*, char*, int, int) (html.cc:3923) ==30983== by 0x80727A9: DilloHtml::write(char*, int, int) (html.cc:531) ==30983== Address 0x661e6bf is 1 bytes before a block of size 16 alloc'd ==30983== at 0x402AC41: malloc (vg_replace_malloc.c:270) ==30983== by 0x80A223F: dw::fltk::FltkPlatform::textToLower(char const*, int) (fltkplatform.cc:584) ==30983== by 0x809E86A: dw::Hyphenator::hyphenateWord(dw::core::Platform*, char const*, int*) (hyphenator.cc:223) ==30983== by 0x809C087: dw::Textblock::hyphenateWord(int) (textblock_linebreaking.cc:762) ==30983== by 0x809BB6E: dw::Textblock::wordWrap(int, bool) (textblock_linebreaking.cc:609) ==30983== by 0x809B6A9: dw::Textblock::processWord(int) (textblock_linebreaking.cc:441) ==30983== by 0x8098753: dw::Textblock::addText0(char const*, unsigned int, short, dw::core::style::Style*, dw::core::Requisition*) (textblock.cc:1694) ==30983== by 0x8097B02: dw::Textblock::addText(char const*, unsigned int, dw::core::style::Style*) (textblock.cc:1478) ==30983== by 0x8073F5C: Html_process_word(DilloHtml*, char const*, int) (html.cc:1216) ==30983== by 0x807A348: Html_write_raw(DilloHtml*, char*, int, int) (html.cc:3923) ==30983== by 0x80727A9: DilloHtml::write(char*, int, int) (html.cc:531) ==30983== by 0x8079D85: Html_callback(int, _CacheClient*) (html.cc:3799)

Dec. 31, 2012

Dpi_connect_socket retry

by jcid＠dillo.org

Hi, On Sat, Dec 29, 2012 at 05:44:22AM +0400, p37sitdu at lavabit.com wrote: > On Thu, Dec 27, 2012 at 04:37:36PM -0300, Jorge Arellano Cid wrote: > > As from the patch you sent (same as before), and the > > explanation you give (ignoring my explanation), I guess you're > > making conclusions out of reading the code in > > Dpi_connect_socket() *only*. > > I made conclusions from reading Dpi_connect_socket() only, because in > the case of error in Dpi_blocking_write file descriptor is left open > and not passed anywhere else. If it was passed somewhere else, I would > read the code there and check if file descriptor was closed there. > > > Have you checked what you state as fact? (i.e. that the FD is > > leaked and not closed by the CCC). FWIW, I did check before > > writing my previous answer. > > Patch that I did for checking is attached. It replaces > Dpi_blocking_write call with -1 to simulate fault in > Dpi_blocking_write for Dpi_connect_socket(). > > Introducing the error in Dpi_blocking_write for all functions doesn't > work, in this case Dpi_connect_socket is not called at all (because > Dpi_start_dpid fails). > > With this patch applied, I recompiled dillo and pressed "bookmarks" > button several times to trigger dpi code. Each time I pressed > "bookmarks" button, new descriptor was leaked: > > % ls /proc/$(pidof dillo)/fd > 0 1 2 3 4 5 > % ls /proc/$(pidof dillo)/fd > 0 1 2 3 4 5 6 > % ls /proc/$(pidof dillo)/fd > 0 1 2 3 4 5 6 > % ls /proc/$(pidof dillo)/fd > 0 1 2 3 4 5 6 7 > % ls /proc/$(pidof dillo)/fd > 0 1 2 3 4 5 6 7 8 > > Then I applied the patch I sent before (with hg qpush) on top of the > fault patch and rebuilt dillo: "make clean && make". I ran dillo again > with ./src/dillo and did the same test with bookmarks button: > > % ls /proc/$(pidof dillo)/fd > 0 1 2 3 4 5 > % ls /proc/$(pidof dillo)/fd > 0 1 2 3 4 5 > % ls /proc/$(pidof dillo)/fd > 0 1 2 3 4 5 > diff -r 8e10859b2aab src/IO/dpi.c > --- a/src/IO/dpi.c > +++ b/src/IO/dpi.c > @@ -641,7 +641,7 @@ > /* send authentication Key (the server closes sock_fd on error) */ > } else if (!(cmd = a_Dpip_build_cmd("cmd=%s msg=%s", "auth", SharedKey))) { > MSG_ERR("[Dpi_connect_socket] Can't make auth message.\n"); > - } else if (Dpi_blocking_write(sock_fd, cmd, strlen(cmd)) == -1) { > + } else if (-1 == -1) { > MSG_ERR("[Dpi_connect_socket] Can't send auth message.\n"); > } else { > ret = sock_fd; Ah, OK, now I see what you mean. There's a semantic ambiguity, the comment: /* send authentication Key (the server closes sock_fd on error) */ means: "if there's an error with the authentication" (I know because I wrote the comment). (and as the patch removed this comment) I made the same test but with: - } else if (!(cmd = a_Dpip_build_cmd("cmd=%s msg=%s", "auth", SharedKey))) { + } else if (!(cmd = a_Dpip_build_cmd("cmd=%s msg=%s","auth",SharedKey+1))) { and the CCC closes the file descriptor as expected. The error case you noticed exists, and although highly unlikely to happen, requires some thought. As from the test above, the socket is connected, then nothing is written to simulate an error, but the OS never sees an error, as it doesn't exist, and doesn't notify either end (IOW, it leaves the socket passively expecting for data to come). Now, the canonical way to handle the whole process should have been to build the CCC upon connect(), then try to authenticate, and let the CCC handle any errors that may come, but this is probably too much code for little gain, so it was all coded in Dpi_connect_socket(). The problem with closing FDs outside the CCC is that it can introduce race conditions, but in this case the CCC is partially built, so your patch suggestion may be OK. I'll have to give it some more thought and tests... -- Cheers Jorge.-

Dec. 30, 2012

ui colors again

by corvid＠lavabit.com

I wrote: > Perhaps it would be good to be able to change colors at runtime, especially > since we already allow the panel size to change. I'll have a look at that > color chooser that fltk has. Here's a quick little toy:

Dec. 30, 2012

show_url pref

by jcid＠dillo.org

On Sun, Dec 30, 2012 at 07:25:59PM +0000, corvid wrote: > > Since it's been broken throughout the dillo3 era, the other day I made it > so that you can't disable it -- and now I'm wondering what the value is in > even having the pref. Removing Location doesn't give us room for anything > more important. It's kind of like having a show_viewport preference. Yes, most probably nobody cares. ;) (nobody has missed it in all these time) Hide_panels is a different thing (much more useful). Probably it'd be good to remove the pref. -- Cheers Jorge.-

Dec. 30, 2012

show_url pref

by corvid＠lavabit.com

Since it's been broken throughout the dillo3 era, the other day I made it so that you can't disable it -- and now I'm wondering what the value is in even having the pref. Removing Location doesn't give us room for anything more important. It's kind of like having a show_viewport preference.

Dec. 30, 2012

ui colors again

by corvid＠lavabit.com

Jorge wrote: > * We may suggest some combinations in the web site or dillorc. > > This way, the traditional look can easily be set, new ones can > be created (suggested) and a default scheme voted. It would be nice if we had a sort of...minimal covering set of the different sorts of color schemes that a person could "sensibly" (whatever that can mean) want for the sake of testing. I imagine the world of graphic design must have such a thing (if only I knew the right words to find it)... Perhaps it would be good to be able to change colors at runtime, especially since we already allow the panel size to change. I'll have a look at that color chooser that fltk has. > Finally, @all, please recieve my best wishes for the new year > that's about to start! ;-) :)

Dec. 30, 2012

ui colors again

by corvid＠lavabit.com

Jorge wrote: > The default theme in dillo is kind of a consensus over the > years. This is, when changes were proposed, the most voted one > (or less objected) got to be the default (same for the website). I think we have a strong tendency to defer to you in general and end up with what Jorge likes, even if that's not your intention particularly. And of course the community is so tiny these days that we often get one vote or zero votes on something, dispiritingly enough. Last year, the opinions presented on the list on ripping out the hardcoded colors were: Benjamin brought it up, iirc, having done so with his fork. I liked the idea. Johannes liked the idea. > * The default colors should not be changed without consensus. It should be quite similar except that there's some white where we customarily have a greenish color. I'm not too crazy about the white, but I didn't want to make it the same greenish merely because it's what we had because then we'd all not notice it and/or accept it as the way it's always been. (The white of course being fltk's default.) I'm thinking that we come up with defaults for foreground and the two backgrounds and maybe selection color, and then put in contrast() and lighter() and darker() as necessary for visibility and pleasant appearance, but base everything on that set of colors instead of hardcoding anything to green or blue or whatever. > * The patch should provide a way to keep the previous default > (i.e. How do I get my colors/pixmaps back?) > > PS: WRT the second point, I find the inactive pixmaps too > "live" for my taste (and thus prefer the pixmaps). I initially wanted to make the inactive pixmaps look better with greyish or darkish backgrounds, and I used Fl_Image's inactive(), which just dims the pixmap down with some background color. But they didn't look inactive enough, so I went the desaturate() route after all, which is rather like using the pixmaps... So I haven't come up with anything that I'm happy enough with to really want to _advocate_ for. > PS2: Is good to have the input/dialog backgrounds configurable > but the defaults should be voted.

Dec. 30, 2012

ui colors again

by jcid＠dillo.org

On Sun, Dec 30, 2012 at 10:52:37AM -0300, Jorge Arellano Cid wrote: > On Sun, Dec 30, 2012 at 03:55:34AM +0000, corvid wrote: > > I wrote: > > > Today I dug up the old ui colors patch from last September, ripped out the > > > part that used system colors as defaults that was a problem*, and played > > > around with it a little. I really like configurable colors, so, since I > > > haven't heard anyone say anything like feature freeze yet, I'll make another > > > push for this thing. > > > > Committed. > > > > - I changed the inactive icon appearance some more. > > - By default it's using the fltk default colors. > > - It might be nice to be able to specify the selection color, too. > > -1 > > The default colors or theme are not easy to define, everybody has > his personal preferences, and there's no single scheme that pleases > everyone. > > The default theme in dillo is kind of a consensus over the > years. This is, when changes were proposed, the most voted one > (or less objected) got to be the default (same for the website). > > This policy works ok for us and is widely used. > > Personally I don't like the change, but I do understand that > other users, may find it fits better. > > So my points are: > > * The default colors should not be changed without consensus. > * The patch should provide a way to keep the previous default > (i.e. How do I get my colors/pixmaps back?) > > PS: WRT the second point, I find the inactive pixmaps too > "live" for my taste (and thus prefer the pixmaps). > PS2: Is good to have the input/dialog backgrounds configurable > but the defaults should be voted. Just trying to make my points a bit clearer, and after playing a bit more with the patch: * +1 to making it configurable (as now) * +1 to add a dillorc option for the second parameter in color_average() call (e.g. setting it to 0.1f makes for the same look as with the handmade pixmaps, so they could be removed). Adding a parameter for the first one may also help (although it's not a priority). * We may suggest some combinations in the web site or dillorc. This way, the traditional look can easily be set, new ones can be created (suggested) and a default scheme voted. Finally, @all, please recieve my best wishes for the new year that's about to start! ;-) -- Cheers Jorge.-

Dec. 30, 2012

CAPI ignores download requests when not in cache or handled by the downloads DPI.

by onepoint＠starurchin.org

I wondered why "Save link as" silently failed on some of my local web pages if I had not already visited the link. It turns out they were "file:///" links and "Save link as" sends a download request which the CAPI silently ignores if the URL is not cached and is not handled by the downloads DPI. I added a MSG_WARN to log this but haven't changed any behaviour. I still found it surprising and was wondering if it could be improved? Regards, Jeremy Henty

Dec. 30, 2012

ui colors again

by jcid＠dillo.org

On Sun, Dec 30, 2012 at 03:55:34AM +0000, corvid wrote: > I wrote: > > Today I dug up the old ui colors patch from last September, ripped out the > > part that used system colors as defaults that was a problem*, and played > > around with it a little. I really like configurable colors, so, since I > > haven't heard anyone say anything like feature freeze yet, I'll make another > > push for this thing. > > Committed. > > - I changed the inactive icon appearance some more. > - By default it's using the fltk default colors. > - It might be nice to be able to specify the selection color, too. -1 The default colors or theme are not easy to define, everybody has his personal preferences, and there's no single scheme that pleases everyone. The default theme in dillo is kind of a consensus over the years. This is, when changes were proposed, the most voted one (or less objected) got to be the default (same for the website). This policy works ok for us and is widely used. Personally I don't like the change, but I do understand that other users, may find it fits better. So my points are: * The default colors should not be changed without consensus. * The patch should provide a way to keep the previous default (i.e. How do I get my colors/pixmaps back?) PS: WRT the second point, I find the inactive pixmaps too "live" for my taste (and thus prefer the pixmaps). PS2: Is good to have the input/dialog backgrounds configurable but the defaults should be voted. -- Cheers Jorge.-

Dec. 30, 2012