- Dillo-dev - lists.mailman3.com

Splitting html.cc: WAS: Re: [Dillo-dev] [patch]: replace form::Form ...

by onepoint＠starurchin.org

On Sun, Jun 01, 2008 at 01:42:11PM -0400, Jorge Arellano Cid wrote: > I removed form.{hh,cc} from cvs. Although they may reappear once > html.cc is split. Well, since you mention it...! :-) I had another go at splitting html.cc and this time I got it to work. No code was modified except that DilloHtml::getCurrentForm is no longer inline and the DilloHtmlForm constructor and destructor are wrapped in Html_form_{create,delete}, which can be called without access to the DilloHtmlForm class definition. So let's talk filenames! I moved the form code into new files called html_form.{hh,cc} . I really like hierarchical file names like that (they make related things sort next to each other), but maybe you would prefer the old form.{hh,cc} ? Also, I had to move a lot of declarations that both html.cc and html_form.cc use into a new header called html_internal.hh . Again, please specify a different file name if you want. I hadn't planned on creating a second new header but there are several definitions that are used by most of the html.cc internals, eg. almost everything needs the DilloHtml class definition and various HT2* macros etc. The "internal header" factors out that common API. I'd like to avoid it but it seems inevitable to me. Note that adding html_internal.hh is a one-off change. Further splitting should not require extra internal headers, just one new header per new source file as is standard. Also note that splitting off the form code is particularly messy since it includes new objects that appear in the DilloHtml definition. I would expect further splitting (eg, the table code) to be much simpler. The final decision is what to do with function names that move from *.cc to *.hh files. Leave them the same or rename "Html_*" to "a_Html_*"? I would suggest leaving them the same because they are still Html internals. They only appear in *.hh files because the internal implementation is spread among several files. If people think this sounds good and confirm the right names for files and functions then I will start proposing patches. I will move the code around in stages without breaking the build, so you'll get several small patches instead of one big one. I think that will be easier to review (but you can have a mega-patch if you really want one). Here are the stats: before: $ wc html*.{hh,cc} 21 42 338 html.hh 0 0 0 html_form.hh 0 0 0 html_internal.hh 6016 20229 191216 html.cc 0 0 0 html_form.cc 6037 20271 191554 total after: $ wc html*.{hh,cc} 21 42 338 html.hh 40 121 1404 html_form.hh 240 841 7832 html_internal.hh 4114 14090 129091 html.cc 1764 5563 57036 html_form.cc 6179 20657 195701 total (The final version may differ as I clean things up, but it won't be much different.) Regards, Jeremy Henty

June 1, 2008

Hooked a decoder for text/plain with charset.

by corvid＠lavabit.com

Jorge wrote: > On Fri, May 30, 2008 at 10:51:33PM +0000, corvid wrote: > > Jorge wrote: > > > On Thu, May 29, 2008 at 09:15:24PM +0000, corvid wrote: > > > > Jorge wrote: > > > > > On Thu, May 29, 2008 at 08:49:22PM +0000, corvid wrote: > > > > > > Jorge wrote: > > > > > > > Just committed this patch. It makes things like: > > > > > > > > > > > > > > http://separat-lib.narod.ru/master.txt > > > > > > > > > > > > > > work. > > > > > > > > > > > > Why now? I don't understand the motivation for this. > > > > > > > > > > > > My next patch is just going to rip it out again because it > > > > > > will become unnecessary. > > > > > > > > > > Yeah, I noticed it could happen after making the patch, > > > > > but a segfault on that page made me fix and commit it. > > > > > > > > Why did it make plain segfault? > > > > > > The patch is attached for you to review in detail. > > > > > > The relevant part is an: > > > > > > Start_Ofs += i - len; > > > > > > substituted by a: > > > > > > Buf_Consumed = BufSize - len; > > > > > > (if the length of a long line got bigger than 'i', 'i - len' > > > was negative. A strange case triggered by that page, that has > > > really long lines). > > > > How did len get bigger than i? > > I tried the page yesterday without being able to get a crash. > > And now I've saved it locally and changed file.dpi to send it > > 200 bytes at a time, which gives me a long series of normal values > > when lines get bigger than packets: > > > > .. > > i-len 76 > > i-len 0 > > i-len 273 > > i-len 112 > > i-len 183 > > i-len 91 > > i-len 0 > > i-len 0 > > i-len 389 > > i-len 101 > > i-len 0 > > i-len 182 > > i-len 0 > > i-len 0 > > .. > > Well, it took me some time to reproduce it again... > > The attached patch is what I've got when adding the decoder over > the old code. It works most of the time, but if you go to: > > http://separat-lib.narod.ru/master.txt > > It will segfault with a Start_Ofs > BufSize. > Note: the patch is against current CVS. Oh, I thought the original no-decoder code was supposed to break.

June 1, 2008

Hooked a decoder for text/plain with charset.

by jcid＠dillo.org

On Fri, May 30, 2008 at 10:51:33PM +0000, corvid wrote: > Jorge wrote: > > On Thu, May 29, 2008 at 09:15:24PM +0000, corvid wrote: > > > Jorge wrote: > > > > On Thu, May 29, 2008 at 08:49:22PM +0000, corvid wrote: > > > > > Jorge wrote: > > > > > > Just committed this patch. It makes things like: > > > > > > > > > > > > http://separat-lib.narod.ru/master.txt > > > > > > > > > > > > work. > > > > > > > > > > Why now? I don't understand the motivation for this. > > > > > > > > > > My next patch is just going to rip it out again because it > > > > > will become unnecessary. > > > > > > > > Yeah, I noticed it could happen after making the patch, > > > > but a segfault on that page made me fix and commit it. > > > > > > Why did it make plain segfault? > > > > The patch is attached for you to review in detail. > > > > The relevant part is an: > > > > Start_Ofs += i - len; > > > > substituted by a: > > > > Buf_Consumed = BufSize - len; > > > > (if the length of a long line got bigger than 'i', 'i - len' > > was negative. A strange case triggered by that page, that has > > really long lines). > > How did len get bigger than i? > I tried the page yesterday without being able to get a crash. > And now I've saved it locally and changed file.dpi to send it > 200 bytes at a time, which gives me a long series of normal values > when lines get bigger than packets: > > .. > i-len 76 > i-len 0 > i-len 273 > i-len 112 > i-len 183 > i-len 91 > i-len 0 > i-len 0 > i-len 389 > i-len 101 > i-len 0 > i-len 182 > i-len 0 > i-len 0 > .. Well, it took me some time to reproduce it again... The attached patch is what I've got when adding the decoder over the old code. It works most of the time, but if you go to: http://separat-lib.narod.ru/master.txt It will segfault with a Start_Ofs > BufSize. Note: the patch is against current CVS. -- Cheers Jorge.-

June 1, 2008

why is SELECT a block container in Tags[]?

by jcid＠dillo.org

On Sat, May 31, 2008 at 07:02:02PM +0000, corvid wrote: > Jorge wrote: > > On Sat, May 31, 2008 at 03:26:50PM +0000, corvid wrote: > > > Why is SELECT a block container rather than an inline container in Tags[]? > > > > Most probably by mistake! > > > > Given that the DTD says: > > > > <!ELEMENT SELECT - - (OPTGROUP|OPTION)+ -- option selector --> > > [...] > > <!ELEMENT OPTGROUP - - (OPTION)+ -- option group --> > > [...] > > <!ELEMENT OPTION - O (#PCDATA) -- selectable choice --> > > > > it looks like inline container would fit better. > > (and stop some segfaults Justus detected). > > Well, it would still provide him with new segfaults for inline stuff like > > <form> > <select> > <font color=red>segfault</font> > </select> > </form> BTW, patch committed (now SELECT is an inline container). With regard to further segfaults, quoting myself: > SELECT can be closed in Html_stack_cleanup_at_open(). SELECT > can contain OPTGROUP and OPTION. > > It would be good to consider the other cases before making a patch. After Justus sends us a set of segfault cases (with the fuzzy technique) we can try to tackle a general solution. -- Cheers Jorge.-

June 1, 2008

[patch]: replace form::Form and remove form.{cc,hh}

by jcid＠dillo.org

On Sat, May 31, 2008 at 05:33:50PM +0100, Jeremy Henty wrote: > > The source files dillo2/src/form.{hh,cc} (which define the form::Form > class) are modified copies of dw2/test/form.{hh,cc} . Dillo uses > almost none of the code in these files. All dillo needs is a Receiver > class that wraps a DilloHtmlForm and delegates calls to the event > handlers to DilloHtmlForm methods. > > This patch removes all references to form::Form and form.{hh,cc} . It > replaces form::Form with DilloHtmlReceiver. It also moves the event > handler code into a DilloHtmlForm method and removes some unnecessary > (void *) casting. This simplifies the code that bridges form events > to DilloHtmlForm methods. > > WARNING: developers must run " make maintainer-clean ; ./autogen.sh " > after this patch appears in their CVS working directory. This is > because automake has to recreate Makefile.in , otherwise the build > will try to compile form.cc which no longer works. > > After this patch is applied you can delete dillo2/src/form.{hh,cc} . > Don't delete dw2/test/form.{hh,cc} though! Committed. I removed form.{hh,cc} from cvs. Although the may reappear once html.cc is split. -- Cheers Jorge.-

June 1, 2008

why is SELECT a block container in Tags[]?

by corvid＠lavabit.com

Jorge wrote: > On Sat, May 31, 2008 at 03:26:50PM +0000, corvid wrote: > > Why is SELECT a block container rather than an inline container in Tags[]? > > Most probably by mistake! > > Given that the DTD says: > > <!ELEMENT SELECT - - (OPTGROUP|OPTION)+ -- option selector --> > [...] > <!ELEMENT OPTGROUP - - (OPTION)+ -- option group --> > [...] > <!ELEMENT OPTION - O (#PCDATA) -- selectable choice --> > > it looks like inline container would fit better. > (and stop some segfaults Justus detected). Well, it would still provide him with new segfaults for inline stuff like <form> <select> <font color=red>segfault</font> </select> </form>

May 31, 2008

why is SELECT a block container in Tags[]?

by jcid＠dillo.org

On Sat, May 31, 2008 at 03:26:50PM +0000, corvid wrote: > Why is SELECT a block container rather than an inline container in Tags[]? Most probably by mistake! Given that the DTD says: <!ELEMENT SELECT - - (OPTGROUP|OPTION)+ -- option selector --> [...] <!ELEMENT OPTGROUP - - (OPTION)+ -- option group --> [...] <!ELEMENT OPTION - O (#PCDATA) -- selectable choice --> it looks like inline container would fit better. (and stop some segfaults Justus detected). -- Cheers Jorge.-

May 31, 2008

[patch]: replace form::Form and remove form.{cc,hh}

by onepoint＠starurchin.org

The source files dillo2/src/form.{hh,cc} (which define the form::Form class) are modified copies of dw2/test/form.{hh,cc} . Dillo uses almost none of the code in these files. All dillo needs is a Receiver class that wraps a DilloHtmlForm and delegates calls to the event handlers to DilloHtmlForm methods. This patch removes all references to form::Form and form.{hh,cc} . It replaces form::Form with DilloHtmlReceiver. It also moves the event handler code into a DilloHtmlForm method and removes some unnecessary (void *) casting. This simplifies the code that bridges form events to DilloHtmlForm methods. WARNING: developers must run " make maintainer-clean ; ./autogen.sh " after this patch appears in their CVS working directory. This is because automake has to recreate Makefile.in , otherwise the build will try to compile form.cc which no longer works. After this patch is applied you can delete dillo2/src/form.{hh,cc} . Don't delete dw2/test/form.{hh,cc} though! Please comment! Regards, Jeremy Henty

May 31, 2008

why is SELECT a block container in Tags[]?

by corvid＠lavabit.com

Why is SELECT a block container rather than an inline container in Tags[]?

May 31, 2008

Hooked a decoder for text/plain with charset.

by corvid＠lavabit.com

Jorge wrote: > On Thu, May 29, 2008 at 09:15:24PM +0000, corvid wrote: > > Jorge wrote: > > > On Thu, May 29, 2008 at 08:49:22PM +0000, corvid wrote: > > > > Jorge wrote: > > > > > Just committed this patch. It makes things like: > > > > > > > > > > http://separat-lib.narod.ru/master.txt > > > > > > > > > > work. > > > > > > > > Why now? I don't understand the motivation for this. > > > > > > > > My next patch is just going to rip it out again because it > > > > will become unnecessary. > > > > > > Yeah, I noticed it could happen after making the patch, > > > but a segfault on that page made me fix and commit it. > > > > Why did it make plain segfault? > > The patch is attached for you to review in detail. > > The relevant part is an: > > Start_Ofs += i - len; > > substituted by a: > > Buf_Consumed = BufSize - len; > > (if the length of a long line got bigger than 'i', 'i - len' > was negative. A strange case triggered by that page, that has > really long lines). How did len get bigger than i? I tried the page yesterday without being able to get a crash. And now I've saved it locally and changed file.dpi to send it 200 bytes at a time, which gives me a long series of normal values when lines get bigger than packets: .. i-len 76 i-len 0 i-len 273 i-len 112 i-len 183 i-len 91 i-len 0 i-len 0 i-len 389 i-len 101 i-len 0 i-len 182 i-len 0 i-len 0 ..

May 31, 2008