- Dillo-dev - lists.mailman3.com

[patch]: ensure Cache_client_make_key() returns > 0
by onepoint＠starurchin.org May 3, 2009

May 3, 2009

Cache_client_make_key() carefully prevents integer overflow from making it return a negative value, but it might return 0. This is a problem because a_Capi_open_url() either calls a_Cache_open_url() and returns the new client key or else returns 0 to signal that it did not call a_Cache_open_url(). If Cache_client_make_key() returns 0 then a_Capi_open_url() will return 0 too, and so its caller will think that a_Cache_open_url() was not called even though it was. This will stop the new cache client from being added to the browser window that is responsible for creating it, which is wrong. Regards, Jeremy Henty

2 1

Style question: cleaning up NULL
by onepoint＠starurchin.org May 3, 2009

May 3, 2009

I've noticed that a_Url_free() and a_Image_unref() both guard against being passed a NULL argument, yet in several cases the code that calls these functions also checks for a NULL, which is redundant. I think that if the API officially accepts NULL then code that calls the API shouldn't be checking for it. Conversely, if the API does not accept NULL then its implementation shouldn't be checking for it. So I would suggest we remove the redundancy by eliminating the NULL check from either the callers or the API implementation. Thoughts? I'll provide patches if people agree and decide whether or not these APIs should support NULL. Regards, Jeremy Henty

2 3

[patch]: fix comment to a_Html_image_new()
by onepoint＠starurchin.org May 2, 2009

May 2, 2009

Does what the subject says. (I think this is my fault: I proposed taking out the "add" argument and renaming the function but forgot the revise the comment.) Regards, Jeremy Henty

2 2

towards a dillo-2.1 release
by jcid＠dillo.org April 28, 2009

April 28, 2009

Hi there, It'd be very good if we can complete these points for the 2.1 release. Listed in no particular order: 1.- Dw documentation: online and as tarball. We have good docs. but to generate them there's a need to have: doxygen, graphviz, dot, latex and friends. That's why having it online and as a tarball is a good idea. 2.- Mapcity segfault In www.mapcity.cl playing with the "Tipo Calle" while the page is loading makes dillo segfault. If "Remote CSS" is disabled, it doesn't happen (also OK in dillo-2.0). 3.- New web site design We have a new banner and after finding a good color palette it may be ready to renew the site before 2.1 is announced. 4.- Debian/Ubuntu packages (after 2.1 final is ready) 5.- Update "8 years" page (it's nine) Yes, I should have done it already... 6.- Configurable key bindings. I'm reviewing this with Tim. Hopefully it will make it into the 2.1 release. 7.- Memory usage statistics Scott has posted some data on this, in another thread. I'll answer there. I'd be working on 5, 6 and 3. Volunteers please assign yourselves to the pending points. BTW, if you have important points to add to the list, please say so. Of course any segfaults or regressions have high priority too. TIA. -- Cheers Jorge.-

4 14

Image callback race condition
by onepoint＠starurchin.org April 28, 2009

April 28, 2009

I am logging the creation and destruction of DilloImage structures (patches attached) and I have confirmed that sometimes the image callbacks run after the Image is freed. Here's a typical example: $ grep '^Image: 0x4687db8: ' ../../Log/log_2009-04-25_21-48-48.txt Image: 0x4687db8: new Image: 0x4687db8: ref: a_Jpeg_callback Image: 0x4687db8: free Image: 0x4687db8: ref: a_Jpeg_callback Image: 0x4687db8: ref: a_Jpeg_callback So how can I track down this race condition? Regards, Jeremy Henty PS: I am still running corvid's img-ref patch (also attached for reference).

2 2

Dillo documentation online
by onepoint＠starurchin.org April 25, 2009

April 25, 2009

There's a tarball of the Dillo documentation at http://starurchin.org/dillo/dillo-docs.tar.gz Alternatively, download the link called "Documentation" at http://starurchin.org/dillo/ It will be updated every night, although it will only be as current as my local HG repo (which I update most days). Regards, Jeremy Henty

1 0

patch: a_Capi_get_flags and following redirection
by corvid＠lavabit.com April 23, 2009

April 23, 2009

When an image on a page wouldn't load due to redirection, I would isolate the image, and often notice that message about Content-Length not matching message body. The reason is in a_Capi_open_url(), where reload is set by checking a_Capi_get_flags() for CAPI_IsCached. It follows the redirection, determines the Location URL isn't cached, and makes another HTTP request for the original URL. The cache entry data for that URL ends up containing the body from the first request with the header and body from the second request appended to it. Ah, but sometimes we want to follow redirection for get_flags. - the download check in capi - the load_stylesheet check in html.cc, assuming that we want to make that work with redirection at some point. - load_now for images in html.cc and the rest don't really matter too terribly much either way. So, unable to think of anything nice or clever, I added an argument to get_flags specifying whether redirection should be followed.

1 0

Instant client-side redirects (META refresh with delay=0)
by jcid＠dillo.org April 18, 2009

April 18, 2009

Hi there, Please excuse me for being silent. Life has poured a few other important things to do over me (e.g. next week I have to move to another apartment). A patch for Instant client-side redirects is fresh in the repo. It took some time to develop because the whole idle/timeout mechanism needs a new API that allows to cancel queued functions. I didn't change the old API (the next release will keep it) but took some time to analyze and it may help with the redraw issue Johannes worked with, and also with hard-to-reproduce crashes. IIRC, Tim asked why several functions trigger from timeouts. Dillo is async. and the main driving function for I/O is Cache_process_queue(). This function is not reentrant, thus timeouts/idle calls allow for it to finish before a new action is taken. In general, there're several parts of dillo that need a more formal working definition or API. They've come to be with the implementation and it'd be good to define and document them, and hence not to have to go through all the involved code to find out how to implement something. I'll review the custom-keys patch series next, and try to re-start the work for the 2.1 release. Please test the patch. I've found it works well with the links in: http://git.kernel.org/gitweb.cgi -- Cheers Jorge.-

1 0

[Patch] add middle_click_closes_tab option fixed
by Christian.Kellermann＠nefkom.net April 12, 2009

April 12, 2009

Skipped content of type multipart/mixed

1 0

[Patch] add middle_click_closes_tab option
by Christian.Kellermann＠nefkom.net April 12, 2009

April 12, 2009

Skipped content of type multipart/mixed

1 1