- Dillo-dev - lists.mailman3.com

Bug in meta tag refresh code [with cause]

by jcid＠dillo.org

On Tue, May 10, 2016 at 08:59:49AM +0200, Johannes Hofmann wrote: > On Mon, May 02, 2016 at 09:59:09PM -0300, Jorge Arellano Cid wrote: > > > [...] > > > > > > I have two pages you can test on - one with the proper meta tag, and one > > > with out: > > > > > > http://irpi.linicks.net:8080/static_simple.html <- good > > > > > > http://fishpi.linicks.net:8081/static_simple.html <- bad - that shows the > > > bug > > > > > > If possible view both with that printf I put in - you can see what mr_url > > > returns. > > > > Please try to be clear & concise. > > > > There's a bug, yes. You may try this: > > > > diff -r c20e74568504 src/html.cc > > --- a/src/html.cc Sun May 01 10:49:17 2016 -0300 > > +++ b/src/html.cc Sun May 01 23:26:48 2016 -0300 > > @@ -3220,6 +3220,8 @@ static void Html_tag_open_meta(DilloHtml > > content = p + strlen("url="); > > else if ((p = strstr(content, ";"))) > > content = p + strlen(";"); > > + else > > + content = ""; > > /* Handle the case of a quoted URL */ > > if (*content == '"' || *content == '\'') { > > if ((p = strchr(content + 1, *content))) > > > > > > @Johannes, does it look OK to you? > > Looks ok to me, but the whole code there is a bit hairy. Yes. This stems from two reasons: the meta-refresh element has no defined standard, and our code has evolved accomodating to work with what we've found along the way. Do you have a testcase battery for meta refresh? Currently, I'm working on the BODY and HTML element handling by the parser (special cases since their implicit optional-magic open and close). Fortunately I have the test cases for BODY and HTML. I also found my old set for META refresh, but I don't have the case you made the last patch for. It would be great to consolidate them. -- Cheers Jorge.-

May 10, 2016

Bug in meta tag refresh code [with cause]

by johannes.hofmann＠gmx.de

On Mon, May 02, 2016 at 09:59:09PM -0300, Jorge Arellano Cid wrote: > Hi, > > On Sun, May 01, 2016 at 04:06:51PM +0100, Nick Warne wrote: > > I have been looking at this for about 5 hours now, and it has fried my > > brain. > > > > Reading the code, and then reading the email thread about depreciated meta > > tag 'refresh', I agree with Dillo's actions. > > > > But what I found was that on some sites (mine included), if there was a tag: > > > > <meta http-equiv="refresh" content="15" /> > > > > Dillo reports and the warning message reports: > > > > The author wanted you to go _here_ after 10 seconds. > > > > What happens now is the link (_here_) gets appended with the refresh value, > > so for example: > > > > www.example.com/refesh.html > > > > becomes: > > > > www.example.com/10 (or 15 or whatever the refresh value is). > > > > I have found that if the meta tag is right: > > > > <meta http-equiv="refresh" content="5;URL='./index.html'" /> > > > > then the code creates the right URL (but it's not seen as no warning is > > given. > > > > html.cc from line 3218 > > > > if the meta tag refresh is correct HTML on the web page, mr_url returns the > > correct path to the refresh. If it isn't, then mr_url returns the value of > > the refresh - which fubars the warning messages link. > > > > I tested this running dillo from the commandline with this insertion and > > testing the correct and wrong meta tags: > > > > } > > /* Skip to anything after "URL=" or ";" if "URL=" is not found */ > > if ((p = dStriAsciiStr(content, "url="))) > > content = p + strlen("url="); > > else if ((p = strstr(content, ";"))) > > content = p + strlen(";"); > > /* Handle the case of a quoted URL */ > > if (*content == '"' || *content == '\'') { > > if ((p = strchr(content + 1, *content))) > > mr_url = dStrndup(content + 1, p - content - 1); > > else > > mr_url = dStrdup(content + 1); > > } else { > > mr_url = dStrdup(content); > > } > > //nick > > printf ("%s\n", mr_url); > > new_url = a_Html_url_new(html, mr_url, NULL, 0); > > // new_url = a_Html_url_new(html, ".", NULL, 0); > > > > > > Not being very good with C++, I can't work out what goes wrong here - but it > > looks like that if 'URL=' or';' isn't found, then the code still looks for > > quotes - which does something to the string? > > > > I have two pages you can test on - one with the proper meta tag, and one > > with out: > > > > http://irpi.linicks.net:8080/static_simple.html <- good > > > > http://fishpi.linicks.net:8081/static_simple.html <- bad - that shows the > > bug > > > > If possible view both with that printf I put in - you can see what mr_url > > returns. > > Please try to be clear & concise. > > There's a bug, yes. You may try this: > > diff -r c20e74568504 src/html.cc > --- a/src/html.cc Sun May 01 10:49:17 2016 -0300 > +++ b/src/html.cc Sun May 01 23:26:48 2016 -0300 > @@ -3220,6 +3220,8 @@ static void Html_tag_open_meta(DilloHtml > content = p + strlen("url="); > else if ((p = strstr(content, ";"))) > content = p + strlen(";"); > + else > + content = ""; > /* Handle the case of a quoted URL */ > if (*content == '"' || *content == '\'') { > if ((p = strchr(content + 1, *content))) > > > @Johannes, does it look OK to you? Looks ok to me, but the whole code there is a bit hairy. Cheers, Johannes

May 10, 2016

image overlaps text

by sgeerken＠dillo.org

Hi Alexander and others, On Mo, Mai 09, 2016, Alexander Voigt wrote: > Dear Sebastian, > > I think one of your recent commits (probably 4543:3cb86de1861a) > introduced a problem with hyphenation. See for example the rendering > of > > debian.org This commit fixes a problem which has hidden another problem which now shows up. I'm working on it. If you want a halfway usable recent version, stick to revision 4542. Sebastian

May 9, 2016

Bug in meta tag refresh code [with cause]

by nick＠linicks.net

On 03/05/16 01:59, Jorge Arellano Cid wrote: > Please try to be clear & concise. I thought I was :) > There's a bug, yes. You may try this: > > diff -r c20e74568504 src/html.cc > --- a/src/html.cc Sun May 01 10:49:17 2016 -0300 > +++ b/src/html.cc Sun May 01 23:26:48 2016 -0300 > @@ -3220,6 +3220,8 @@ static void Html_tag_open_meta(DilloHtml > content = p + strlen("url="); > else if ((p = strstr(content, ";"))) > content = p + strlen(";"); > + else > + content = ""; > /* Handle the case of a quoted URL */ > if (*content == '"' || *content == '\'') { > if ((p = strchr(content + 1, *content))) > > > @Johannes, does it look OK to you? OK, that doesn't quite work correctly - it removes the warning messages altogether. Using: content = "./"; produces the warning message and with the top level URL - but, a case like this: www.example.com/refresh.html should produce a refresh to that URL, but it drops the page and points to www.example.com/ Nearly there... Nick -- Gosh that takes me back... or is it forward? That's the trouble with time travel, you never can tell." -- Doctor Who "Androids of Tara"

May 3, 2016

"ok" for unnested body

by jcid＠dillo.org

Hi, On Sat, Apr 30, 2016 at 11:48:18AM -0300, genoob wrote: > Hello! > I'm a noob, so I'm sorry if this is actually not an issue, but I've > searched and haven't found anything on this. > > When creating a simple list using `ul` (I've replaced < > with `` in > order to avoid html parsing by some readers) tags I accidentally > placed it after `/body``/html` and to my surprise It wasn't marked as > an html error and the data was correctly displayed, jus like it would > be had the list been placed inside the `html``body` `/body``/html` > tags. Unnesting `li` from `ul` however displayed nesting problem, but > only referring to the `li` being out of `ul` or `ol` tags, not from it > being unnested from `html` `body`. > > I realize my explanation might be a little confusing at this point, so > this is what I mean: > > I put it on bpaste so this message wouldn't be flagged as spam: > https://bpaste.net/show/ed1374304997 > And this is a screenshot of the data being displayed and getting a > thumbs-up from the browser in case its relevant: > http://i.imgur.com/d1f9Pv9.png > > I'd like to report this in case it's a bug(invalid html being marked > valid), however if it's a valid html you can disregard/exclude this > message from the list with no hard feelings. > > I'm using dillo-3.0.5-r1 by the way. > > I'm sorry if this was confusing and wasted your time. > In case you need more info on my system I'm happy to help however I can. > > And thanks for maintaining such a nice project (It really helped me a > lot thus far). A bit of history... Long ago, dillo stopped rendering after BODY or HTML was closed. Unfortunately several free web hosting sites were serving their client's sites wrapped within their formats (mainly advertising); fair enough. But they were using poor HTML (tag soup) and used to close BODY even more than twice. Dillo users complained that their sites were not visible with Dillo... At some point we decided to keep rendering until EOF. Now, coding detection for this type of bug is possible and not that complex, it is just low on priority (i.e. patch welcomed). Anyway, it needs a solid understanding of the parsing code. I'll give it a quick look and if I can come with a short patch, it could be committed. FYI, if you right click the bugmeter you can validate the page either with W3C or WDG (more info in the same menu). -- Cheers Jorge.-

May 3, 2016

Bug in meta tag refresh code [with cause]

by jcid＠dillo.org

Hi, On Sun, May 01, 2016 at 04:06:51PM +0100, Nick Warne wrote: > I have been looking at this for about 5 hours now, and it has fried my > brain. > > Reading the code, and then reading the email thread about depreciated meta > tag 'refresh', I agree with Dillo's actions. > > But what I found was that on some sites (mine included), if there was a tag: > > <meta http-equiv="refresh" content="15" /> > > Dillo reports and the warning message reports: > > The author wanted you to go _here_ after 10 seconds. > > What happens now is the link (_here_) gets appended with the refresh value, > so for example: > > www.example.com/refesh.html > > becomes: > > www.example.com/10 (or 15 or whatever the refresh value is). > > I have found that if the meta tag is right: > > <meta http-equiv="refresh" content="5;URL='./index.html'" /> > > then the code creates the right URL (but it's not seen as no warning is > given. > > html.cc from line 3218 > > if the meta tag refresh is correct HTML on the web page, mr_url returns the > correct path to the refresh. If it isn't, then mr_url returns the value of > the refresh - which fubars the warning messages link. > > I tested this running dillo from the commandline with this insertion and > testing the correct and wrong meta tags: > > } > /* Skip to anything after "URL=" or ";" if "URL=" is not found */ > if ((p = dStriAsciiStr(content, "url="))) > content = p + strlen("url="); > else if ((p = strstr(content, ";"))) > content = p + strlen(";"); > /* Handle the case of a quoted URL */ > if (*content == '"' || *content == '\'') { > if ((p = strchr(content + 1, *content))) > mr_url = dStrndup(content + 1, p - content - 1); > else > mr_url = dStrdup(content + 1); > } else { > mr_url = dStrdup(content); > } > //nick > printf ("%s\n", mr_url); > new_url = a_Html_url_new(html, mr_url, NULL, 0); > // new_url = a_Html_url_new(html, ".", NULL, 0); > > > Not being very good with C++, I can't work out what goes wrong here - but it > looks like that if 'URL=' or';' isn't found, then the code still looks for > quotes - which does something to the string? > > I have two pages you can test on - one with the proper meta tag, and one > with out: > > http://irpi.linicks.net:8080/static_simple.html <- good > > http://fishpi.linicks.net:8081/static_simple.html <- bad - that shows the > bug > > If possible view both with that printf I put in - you can see what mr_url > returns. Please try to be clear & concise. There's a bug, yes. You may try this: diff -r c20e74568504 src/html.cc --- a/src/html.cc Sun May 01 10:49:17 2016 -0300 +++ b/src/html.cc Sun May 01 23:26:48 2016 -0300 @@ -3220,6 +3220,8 @@ static void Html_tag_open_meta(DilloHtml content = p + strlen("url="); else if ((p = strstr(content, ";"))) content = p + strlen(";"); + else + content = ""; /* Handle the case of a quoted URL */ if (*content == '"' || *content == '\'') { if ((p = strchr(content + 1, *content))) @Johannes, does it look OK to you? -- Cheers Jorge.-

May 3, 2016

CPU hog bug with https?

by sgeerken＠dillo.org

On So, Mai 01, 2016, Jorge Arellano Cid wrote: > Hi, > > If I start current dillo with [1], an https dialog springs > (untrusted certificate). Answering "continue" gets into a CPU hog, > answering "cancel" allows reading the page with no problems. > > FWIW, I've done that lots of times with other sites, and this is the > first CPU hog I get accross. > > > [1] https://linux.slashdot.org/comments.pl?sid=5852295&cid=48188823 I can reproduce this. However, profiling shows rather problems in the rendering code (dw::core::Widget::queueResize etc.), so the problem is probably not related to HTTPS. Also, mirroring this page ("wget -pkH") and displaying it locally shows the same symptoms. Looks that this belongs rather on my todo list. Sebastian

May 1, 2016

CPU hog bug with https?

by eocene＠gmx.com

On Sun, May 01, 2016 at 12:36:39PM -0300, Jorge Arellano Cid wrote: > If I start current dillo with [1], an https dialog springs > (untrusted certificate). Answering "continue" gets into a CPU hog, > answering "cancel" allows reading the page with no problems. > > FWIW, I've done that lots of times with other sites, and this is the > first CPU hog I get accross. > > > [1] https://linux.slashdot.org/comments.pl?sid=5852295&cid=48188823 I'm not getting a dialog here. What is the page pulling in that dillo is complaining about? I allowed linux.slashdot.org www.gstatic.com linux.slashdot.org analytics.slashdotmedia.com linux.slashdot.org a.fsdn.com linux.slashdot.org s3-us-west-2.amazonaws.com in domainrc, but it doesn't look like any of those are over https.

May 1, 2016

CPU hog bug with https?

by jcid＠dillo.org

Hi, If I start current dillo with [1], an https dialog springs (untrusted certificate). Answering "continue" gets into a CPU hog, answering "cancel" allows reading the page with no problems. FWIW, I've done that lots of times with other sites, and this is the first CPU hog I get accross. [1] https://linux.slashdot.org/comments.pl?sid=5852295&cid=48188823 -- Cheers Jorge.-

May 1, 2016

Bug in meta tag refresh code [with cause]

by nick＠linicks.net

I have been looking at this for about 5 hours now, and it has fried my brain. Reading the code, and then reading the email thread about depreciated meta tag 'refresh', I agree with Dillo's actions. But what I found was that on some sites (mine included), if there was a tag: <meta http-equiv="refresh" content="15" /> Dillo reports and the warning message reports: The author wanted you to go _here_ after 10 seconds. What happens now is the link (_here_) gets appended with the refresh value, so for example: www.example.com/refesh.html becomes: www.example.com/10 (or 15 or whatever the refresh value is). I have found that if the meta tag is right: <meta http-equiv="refresh" content="5;URL='./index.html'" /> then the code creates the right URL (but it's not seen as no warning is given. html.cc from line 3218 if the meta tag refresh is correct HTML on the web page, mr_url returns the correct path to the refresh. If it isn't, then mr_url returns the value of the refresh - which fubars the warning messages link. I tested this running dillo from the commandline with this insertion and testing the correct and wrong meta tags: } /* Skip to anything after "URL=" or ";" if "URL=" is not found */ if ((p = dStriAsciiStr(content, "url="))) content = p + strlen("url="); else if ((p = strstr(content, ";"))) content = p + strlen(";"); /* Handle the case of a quoted URL */ if (*content == '"' || *content == '\'') { if ((p = strchr(content + 1, *content))) mr_url = dStrndup(content + 1, p - content - 1); else mr_url = dStrdup(content + 1); } else { mr_url = dStrdup(content); } //nick printf ("%s\n", mr_url); new_url = a_Html_url_new(html, mr_url, NULL, 0); // new_url = a_Html_url_new(html, ".", NULL, 0); Not being very good with C++, I can't work out what goes wrong here - but it looks like that if 'URL=' or';' isn't found, then the code still looks for quotes - which does something to the string? I have two pages you can test on - one with the proper meta tag, and one with out: http://irpi.linicks.net:8080/static_simple.html <- good http://fishpi.linicks.net:8081/static_simple.html <- bad - that shows the bug If possible view both with that printf I put in - you can see what mr_url returns. Phew* Nick -- Gosh that takes me back... or is it forward? That's the trouble with time travel, you never can tell." -- Doctor Who "Androids of Tara"

May 1, 2016